This change modifeis Options.hpp to add the possibility of
vector-valued options to the config file and command line.
From the 'case' code, this option is used by (for example)

std::vector<int> intvec;
add_option("Name",&intvec,"Description);

When run, options of the form --Name 1 2 3 4 (at the command
line) or Name = 1 2 3 4 (in the config file) will be added
to the vector 'intvec' as separate entries.

Because add_option is a template function, this works
identically for double (floating point)-valued options.  It
has not been tested with string options.

Since an empty vector is a perfectly cromulent and
unambiguous possibility, vector-valued options do *not*
permit the specification of defaults.

This modifies the helper make_deps.sh script to download
Boost 1.6.1, rather than the previous 1.5.1 version.  At
least for now, the URL for the download points to
Sourceforge rather than csubich's public-facing web presence
on belize.math.uwaterloo.ca.

If Sourceforge shuts down, or if Boost moves its historic
releases off the site, then this will need modified.

After some experimentation, the previous commit of MPI-IO file
output/input had (and has) lingering bugs.  To prevent breaking existing
code, these changes have been moved into a 'testing' branch and the
'master' branch has been reverted to its prior state with this merge.

Once the MPI-IO output is fully working, it will be restored into the
master branch.

- Added 1D and 2D outputs
- Added parallel writing and reading

The automatic_grid() helper function in BaseCase was not properly
functional if called without supplied arguments for the 1D grids
xx/yy/zz.  Now, if these arguments are null or not supplied, the
function will do its grid generation using temporarily-allocated
internal 1D arrays, deleted at the end of its execution.

Previously, the automatic_grid function was creating Chebyshev-type
grids with points ordered as:

{x,z} = L/2 + L/2*cos(pi*ii/(N-1))

This is incorrect.  The expected ordering is actually the reverse, or

{x,z} = L/2 - L/2*cos(...)

and write vertical chain was moved to BaseCase (can write out vertical profile at a high temporal frequency)

Safety dump estimates output write time and writes fields with enough time before the computation limit is reached

Parformer.cpp -- previously, an all-NULL transform being
asked for its wavenumbers would only return an array of size
one, rather than one of appropriate length for the (real)
"temporary" array.  This could conceivably have caused
problems for the graceful use of 3D routines on
2D-but-layered fields.

The pair of Transformer.[ch]pp files contained old code built for SPINS
before parallel transformation (over MPI) was fully implemented.  These
classes were replaced wholesale with Parformer.[ch]pp (Parallel
Transformer), but the older code remained in the repository.

Removing these files should not affect any active code.  It will affect
certain disued test cases (subject to changing #includes), but these
cases themselves need more serious maintenance to update for the
"gradient object" interface.

At startup, print a string describing the dimension
type and spectral expansion type, rather than the
internal enum value.

The matlab readers incorrectly interpreted 58 as ':', which resulted it in returning a vector rather than a scalar when index 58 is requested.

Previously for the wave_reader case file, "tracer.0" and
"tracer_reader.m" files were being output regardless of whether the
configuration specified that there was a passive tracer.  This arose
because of a scoping error during tracer initalization; tracer.0
contained a copy of rho.0.

This has been fixed in this commit.  No numerical results were in error
because of this bug.

The case file source code is embedded into the binary at compile time.
At runtime, the source code is printed to spinscase.cpp. This eliminates
the need to maintain a copy of the case file alongside the binary.

The quadrature weight "get" functions will now check whether the
corresponding array has been initialized (length >= 1) before returning
the array.  This will catch errors where quadrature weights are used
before being initialized.

The "automatic" (unmapped) grid calculation used by wave_reader
(automatic_grid in BaseCase.cpp) was in error for grids of
Chebyshev-type.  Instead of computing:

z = -Lz/2 + Lz/2*cos(pi*(0:(N-1))/(N-1))

it instead computed

z = -Lz/2 + Lz/2*cos(pi*(0:(N-1))/N)

(to use matlab-style notation).  This caused the written-out zgrid to be
in error near the bottom (max z-index) boundary.  Files using a
chebyshev-type unmapped x-coordinate would see the error as well in the
x-grid (max x-index).

This error affects only processing that uses the grid variables itself.
Within wave_reader where this was used, the automatic-generated grids
were used only for writing out; the input velocities and density were
read directly in from their corresponding files.  Likewise, SPINS
performed the "correct" differentiation internally without reference to
the grid coordinate.

Also recovered fingers_tank from the old case files, as it
functions as a decent introduction to double-diffusive
instability.

As compiler versions march on, some previously-valid code was generating
warnings; these have been corrected.

*) In Plan_holder and T_util, using <ext/hash_map> was generating a
deprecation warning.  The proper replacement for the GCC <hash_map> is
the STL <unordered_map>, however that is part of the C++-0x standard
that is not yet supported by default in GCC 4.3-4.6, requiring a
compile-time command-line flag.

Since implementing *that* over a broad range of systems will be
extremely difficult, the better option for now is to silence the
deprecation warning by selectively undefining __DEPRECATED around the
inclusion of <ext/hash_map>.  This preserves any GCC deprecation
warnings around other bits of code.

*) In T_util and gmres_1d_solver, "literal string constants" were being
assigned to char *, when string literals are supposed to be const char
*.  This reflected an error in the variable spec; there's no reason that
the assigned-to-vars couldn't themselves be const char *.

As of this commit, the code compiles (make cases/bench_dipole.x)
compiles warning-free on belize.

The old dipole cases in the cases/ directory (dipole crashing into
no-slip boundary) were initially written before the 'analysis'
routine was given pressure.  Adding pressure-analysis to NSIntegrator
changed the method's prototype, leading to a compile-time error on
these files.  The fix is to add a dummy 'pressure' reference to the
dipoles' analysis methods; the behaviour of the other parameters is
unchanged so no logic changes are necessary.

This update adds an explicit MPI Barrier to the end of the split 2D file
reader.  This should address a race condition discovered by Nancy when
resuming a 2D case as 3D on Scinet, where the initially-2D files were
(correctly) overwritten by their 3D versions.  It appears that the
overwrite happened on some processes before others had opened the file,
causing the later processes to read incomplete files.  The exact
sequence appears to have been:

* Process 1 opens 'u.0' as a 2D-split file to read
* Process 1 finishes its reading and assigns the contents to its
  internally-3D array
* Process 1 begins writing out its segment of the 3D array, as 'u.0'.
  (It blocks on the write-synchronization, but has already written to
  disk.)
* Process 2 wakes up (on a remote node) and opens 'u.0' as a
  presumably-2D file.  It dies.

This race condition could, under different circumstances, also cause
data corruption (if process N wasn't reading past the end of the file in
a linear sense; possible for large Ny and small Np).  The corrective
measure was to apply an MPI Barrier at the end of read_2d_slice in
Split_reader_impl.cc (the implementation fucntion used by the more
user-friendly functions in Science.cpp).

This patch specifically includes -ftz for winisk builds; this
forces denormalized floats to be flushed to 0, which should
avoid software-floating-point-faults that spam /var/log/messages
and probably slow down the program.  (Quick checks suggest a ~8%
speed increase for bench_dipole).

Additionally, make_deps is also extended to specifically include
CFLAGS and mark building with CC on the ./configure command line.
This should ensure that the library is built with the same
compiler (and settings) as SPINS itself.  If you have to modify
this part, pay special attention to the comment abour use of bash
array syntax.

Prior modifications to systems/plata.sh did not result in a clean
build; spins itself and the UMFPACK libraries were built in 64-
bit with the -m64 option, but fftw, blitz, and boost were not.
This led to build-time errors.  The fix involves making sure that
make_deps specifies CFLAGS/CXXFLAGS as appropriate, and this was
done via including $BOOST_OPTIONS/$FFTW_OPTIONS/etc flags.  Where
those didn't already exist in make_deps, they were added so that
it will be easy to make a similar change in the future for other
systems.