The Date Repeat module’s date_repeat_build_dates() function does not support
RDATEs and EXDATEs defined as separate properties, as well as being a bit
buggy in other ways. So I copy-pasta’d it and made the necessary fixes.

I wanted to add support for EXRULEs, but it’s more complicated than I
expected. So since no one has actually asked for it, I left it as a TODO.

Date iCal 3's use of the Libraries APIs was slightly less awful than 2's, but it
would still fail to properly detect an incorrectly installed iCalcreator.

This is a forward-port of the fix that I just submitted to the 2.x branch.

Because OSX uses a case-insensitive file system, git didn't pick up the change
I made when I re-wrote the parser. In retrospect, I probably should have chosen
a totally new name for the feeds parser, but the existing filename *was* the
most appropriate name for the new class, just with different capitalization.

As jakemonO pointed out, the Calendar will not display All Day events which have no
end date value, which means that Date iCal's method of leaving off the end date for
single-day All Day events was a really bad idea. Fortuantely, setting the end date
to the same day as the start date makes the date node display correctly for single-
day events, which was the reason that the end date was being left off in the first
place. This means that every All Day event can simply have it's end date rewound by
one day, and all is well in both the node display and the Calendar.

User oskar_calvo discovered that if you update Date iCal from 1.x, but
don't also install the Libraries module, you'll get a fatal error when Date iCal
attempts to check for the iCalcreator library, because the libraries_detect()
function doesn't exist.

Now, you'll instead see an error in the site requirements.

Rather than defaulting an an illegal value, the Fields plugin's settings
form now offers "First populated Date field" as the default option for
"Date field". Using this option will tell Date iCal to look through all the
fields in the View until it finds a populated Date field, and use that field.

So, if there are two Date fields specified in the view's FIELDS setting, and
at runtime the value of the first one is empty (e.g. because the node
doesn't use that field), then the second Date field's value will be used.

The update hook now also checks for importers which are set to the new parser,
but still have un-capitalized source keys. This will help users who forget to
run the database update before manually fixing the missing
DateIcalIcalcreatorParser warning.

This commit adds the migration code that converts all Feeds Importers which used
the DateIcalIcalcreatorParser plugin to use DateiCalFeedsImporter instead. The
old classes have been removed.

Renamed hook_date_ical_import_calendar_alter to hook_date_ical_import_vcalendar_alter, to match the export hook.

Also did some minor refactoring and comment editing.

Date iCal 3.0 is nearing completion! The only major task remaining
is to write the migration code.

iCal feeds are now parsed in batches of 50 (by default).

The Date parser code has been significantly improved, removing the
old limitation of needing to parse DTSTART first, and also covering
more edge cases and error conditions.

I hadn't yet tested the RRULE import code when I first pushed Date iCal 3.x-dev,
but now I have. It's much more robust than it was in 2.x.

In addition, I moved the parsing class out of DateiCalFeedsParser.inc and into
libaries/ParserVcalendar.inc, which is like how Feeds' own CSV parser is set up.

This new version is still in a state of partial completeness. You may need to do
some manual fixing to get it working, due to a change I made in the classes
offered by this module. Try 'drush cc registry' if 'drush cc all' fails.

Please note that ALL of the hooks exposed by Date iCal have changed. They do
essentially the same things they used to do, but they're named differently.

The big change is that the Feeds plugin has been completely re-written to
conform to the Feeds APIs. It's now much more resilient and powerful. Feeds
Tamper will now work, and is in fact the prescribed way to alter data, since I
removed several alter hooks.

Some ical feed readers will erroneously take the DTSTAMP field into
consideration when comparing a new download to a previous one, and end up
treating all events as having been updated, because their DTSTAMP field is
different. This new option lets you get around those buggy readers.

I also added the CREATED field, which was embarassingly missing until now.

LAST-MODIFIED was only being included if $entity->changed was set, and there
may have been some cases where that wasn't true. Now, LAST-MODIFIED will be
included based on $entity->created, if $entity->changed isn't set.

Much like in issue [#1935820], Date's bizarre implementation for repeating date
values was causing every VEVENT with an RRULE to be duplicated for each repeat.
I fixed the Fields plugin the same way that vlad fixed Entity.

According to the spec, EXDATE properties can have multiple values. However, many
popular calendar client don't support that, and only accept iCal feeds which
use multiple separate EXDATE properties. That is spec-compliant, though, so I
refactored Date iCal to do that for EXDATEs and RDATEs.

The previous fix that I applied for issue #2070883 was the complete wrong way to
go about fixing that problem, because it introduced a new, much more sinister
issue: the UID for an event would change every time the iCal feed got downloaded.

According to the spec, the UID is supposed to be persistent, so I re-engineered
the algoithm that creates UIDs for the Fields plugin so it will create the same
UIDs as the Entity plugin does.

iCalcreator requires that you *not* call createTimezone($vcal, 'UTC'), because
the spec says to use zulu time rather than a UTC VTIMEZONE object.

Also fixed a copy-pasta bug that probably never manifested, since it would have
only had an effect on events that have an end date that's in a different
timezone from the start date.

Also updated the README to reference the iCalcreator GitHub.