The Identify method had two bugs when processing fdAT (FrameData) chunks:
1. A spurious Skip(4) before SkipChunkDataAndCrc caused the stream to
be misaligned by 4 bytes, since chunk.Length already includes the
4-byte sequence number.
2. Unlike Decode, which consumes all fdAT chunks for a frame in one shot
via ReadScanlines + ReadNextFrameDataChunk, Identify processed them
individually, calling InitializeFrameMetadata for each chunk and
inflating the frame count.
The fix removes the extra Skip(4) and adds SkipRemainingFrameDataChunks
to consume all continuation fdAT chunks for a frame, mirroring how
ReadNextFrameDataChunk works during decoding.
Andreas Eriksson
ac1905328d