Our SpriteFrameType names refer to the byte channel order rather than
the bit order, meaning that SpriteFrameType.BGRA corresponds to the
standard Color.ToArgb() etc byte order when the (little-endian) integer
is read as 4 individual bytes.
The previous code did not account for the fact that non-indexed Png
uses big-endian storage for its RGBA colours, and that SheetBuilder
had the color channels incorrectly swapped to match and cancel this out.
New SpriteFrameType enums are introduced to distinguish between BGRA
(little-endian) and RGBA (big-endian) formats, and also for 24bit data
without alpha. The channel swizzling / alpha creation is now handled
when copying into the texture atlas, removing the need for non-png
ISpriteLoader implementations to allocate an additional temporary array
and reorder the channels during load.
Passing the getUsedFrames argument will cause frames to be buffered
locally (and not added to the SheetBuilder) until they are explicitly
requested. This allows unused frames in sprite files to be skipped
instead of wasting space in GPU memory.
We split the caching SpriteLoader into a SpriteCache and FrameCache. SpriteLoader instead becomes a holder for static loading methods.
Only a few classes loaded sprite frames, and they all use it with a transient cache. By moving this method into a new class, we can lose the now redundant frame cache, saving on memory significantly since the frame data array can be reclaimed by the GC. This saves ~58 MiB on frames and ~4 MiB on the caching dictionary in simple tests.
