Fix incorrect dealing with offset sub-frames

[Jimbly]

gifwrap fails to correctly decode test/fixtures/rnaples-offsets-public.gif included in its tests. I threw together a simple test that combines all frames into one image with Jimp.

Erroneous results:

Expected results (after applying fix):

omggif's decodeAndBlitFrameRGBA blits the entire frame, not just the sub-region, so, in order to keep the GifCode API the same, we must contract the buffer, skipping all of the empty space.

An alternative solution would be to simply set info.[width/height] = reader.[width/height]; info.[x/y] = 0, which would be slightly faster in here (but cause clients to be slower, as they have to blend the whole image).

Better yet would be to fix omggif (or use a different library) to allow returning just the sub-section, however I looked into decodeAndblitFrameRGBA and it gets pretty complicated due to interlacing and stuff.

But, for a quick fix, this'll do.

[devinivy]

I can't verify the fix itself, but this solves similar problems I was having with an animated gif as well.

[Jimbly]

Yeah, without this fix, I seem to recall that any animated GIFs that animated a sub-frame (don't replace the entire frame each time) won't work.

[devinivy]

Ah! Yeah, that is consistent with what I've seen. At least the glitch looks neat :)

[jtlapp]

Thank you sooo much for finding and fixing this problem. For now, we'll call it expected interface protocol for Omggif. I also added a test to confirm expected behavior.

[Jimbly]

Glad to help, thanks for merging!