Major refactoring plan

[bvibber]

Been in need of a major refactor for some time. May still take some time to get to. Rough plan:

APIs and internals:

• provide an AVAssetReader and AVAsserWriter-like API for decode and encode; the current OGVDecoder API is too confusing to use and if anything goes wrong it's unclear what happened. Add a seek method on top of the track readers.
• always copy to sample buffers for both video and audio decodes, instead of playing locking/lifetime games with the buffers
• use AVSampleBufferDisplayLayer and drop OGVFrameView entirely (upstream blockers fixed in iOS 11)
• decouple codecs from demuxer in the decode side as well as the encode side
• allow each codec and de/muxer module to register itself with core at init time
• drop the old AVAssetReader-based MP4 reader (limited to local files, no URLs, no support for Opus/VP9/etc) and replace with a proper demuxer. Use VideoToolbox or whatever is current for decoding H.264.

Build system:

• drop CocoaPods requirement (if end up publishing as a pod, do it with the framework output)
• have a single Xcode project that generates multiple frameworks
  • OGVKit - core framework with no muxers or codecs
  • OGVVorbis, OGVOpus - frameworks wrapping audio codecs
  • OGVTheora, OGVVPX, OGVAOM (todo) - frameworks wrapping video codecs
  • OGVOgg, OGVWebM, OGVMP4 (todo) - frameworks wrapping demuxers/muxers
• each of the format-specific frameworks registers itself with the OGVKit core at init time
• build all format libraries (not just VPX) with their own build systems, allowing asm for all supported platforms
• statically link the format libraries in, so you only need the top-level frameworks