This is needed for blocked layout, as it can't do normal interleave.
Probably could be fixed in the future to remove several superfluous
_int/block decoders
Previously, the streams' offsets needed to be pre-adjusted with the
interleave; now assumes all streams start in the same offset (first
stream).
This simplifies short last interleaves (SCD/P3D) and makes
layout_mpeg_custom unnecessary (also allows theoretical variable-sized
interleaves).
The layout was designed to do subframe deinterleave (when
less-than-a-frame bytes of each channel are interleaved) in an array and
pass it to "mem" decoders.
In practice this only happens in a handful of formats, was only used
with DSP, and since making "mem" decoders is required it's simpler to
make normal decoders handling the byte layout directly.
In an effort to simplify vgmstream's layouts, code, and other esoteric
features I changed the old decode_ngc_dsp_mem for decode_ngc_dsp_subint;
results are byte-exact.
EACS was just DVI (high nibble first) with stereo and mono modes, while
old DVI was mono only.
This unifies both decoders, so DVI_IMA (not interleaved) works with mono
and stereo while DVI_IMA_int (interleaved) forces mono.
Some metas needed to explicitly set DVI_IMA_int but others work with no
change.
Those are custom read/seeks that can transform data on the fly, for
games that use variations of regular FFmpeg codecs. When FFmpeg tries to
read N bytes, the code may read less/more but still fill the buffer with
N transformed bytes. Must handle "real" (data) and "virtual"
(transformed) offset/sizes.
Works with fake headers too, but "virtual" offset/size now include the
fake header, as I found hard to keep in mind when the header_size was
being added/substracted (could be clearer).
