Merge pull request #1010 from rouault/subtile_decoding_stage3

Subtile decoding: memory use reduction and perf improvements

diff --cc src/bin/jp2/opj_compress.c
Simple merge

diff --cc src/lib/openjp2/j2k.c
Simple merge

diff --cc src/lib/openjp2/openjpeg.h
Simple merge

diff --cc src/lib/openjp2/tcd.c
Simple merge