ASTC weights SIMD encoding

ssse3 (I5 6300) : 163 => 136 ms
arm (A53) : 340 => 282 ms
I moved the block weight transform code in a single function : pack_astc_block_weights
you can enable/disable the SIMD code with a define
BASISD_ASTC_SIMD
All the simd code is annotated

tested x86_64 on windows, compiled with VS2019 and clang 11
tested arm & arm64 on android, compiled with latest NDK (clang11)

if needed, you can get in touch at :
ronan.bel@gmail.com
ronan.bel@ubisoft.com
fix previous issue + optimize unpack

Author: ronanbel