Just wanted the ability to be able to add private data to the properties
data.  Makes it a little easier to manage data if you get updates from
controls.

The pointer here is confusing, so I'm just going to remove it and have
it return the value instead.

Also avoid allocation roundtrip if the file is empty after the BOM

Ideally this should populate the list with matching OBS samplerstate_ts
like the GL code, but unfortunately there is no mapping for that yet

On OSX clicking the X title bar button immediately destroys "all" native
windows (after sending a close event) which causes
[NSSurface _disposeSurface] to crash if invoked while GL is using
the surface

This should fix the previews in most cases when moving between monitors
with differing pixel ratios

This should fix GS rendering on surfaces on HiDPI displays; moving
windows between displays with differing pixel ratios currently requires
a manual resize

Move from glLoadGen to GLAD

Previously the properties window would sometimes not receive
a closeEvent, leaving a dangling pointer in OBSBasicMain (and resulting
in a crash when trying to open a new properties window or closing
the application)

kCGDisplayStreamSourceRect takes points rather than pixels

This automagically creates a (localized?) Preferences entry on OSX as well

My prior code was incorrect;  I mixed up the two parameters, the
GL_TEXTURE_SWIZZLE_* parameter specifies the target channel, and the
value itself specifies the source channel.,  If that makes sense.

Avoids problems if swizzles are used elsewhere (like in the newer X11
plugins)

VC2013 still does not properly support placement of variables anywhere
in the scope.  It's a garbage compiler, always will be a garbage
compiler.

Conflicts:
	test/osx/test.mm

The test program and test filter wasn't working properly because the ID
for it had actually changed.

The test programs also weren't updated for the new main render callbacks
which must be used when making a program.

Also remove an unused variable from obs-encoder.c (via clang static
analysis)

This also makes the clang static analyzer happy (it complained about
uninitialized fields in the AbsoluteTime struct)

For the strangest reason, it would cause errors when linking to ffmpeg
on the 32bit windows release build

Packets were not interleaving, thus new data was being sent out with
potentially non-monotonically increasing timestamps

Improve bundle fixup

Before, async video sources would flicker because they were only being
drawn when they were updated.  So when updated, they'd draw that frame,
then it would stop drawing it until it updated again.  This fixes that
issue and they should now draw properly.

Also, fix a few other minor bugs and issues relating to async video,
and make it so that non-async video filters can be properly applied to
them.

For the purposes of testing, change the 'test-random' source to an async
video source that updates every quarter of a second with a new random
face.

Also fix a bug where non-async video sources wouldn't have filter
effects applied properly.

Was wrong to use the swizzle -- it was an incorrect assumption on my
part that this could work in the first place (despite working on
windows)

Linux fixes and additions