[Draft] Opencl lz4

[kif]

Changelog: LZ4 compressor on GPU

• OpenCL code targeted to run on GPU
  • Find position of interest
  • Create segment descriptor (position in input, length of litterals and matches)
  • Concatenate several descriptor (between passes, in the case there are only litterals)
  • Write compressed stream, including tokens, offsets, overflows in litteral and matches
  • Manage the beginning of stream (4 bytes with uncompressed buffer size)
  • Manage end of stream (cannot end with matches, has to finish with litterals
  • Concatenate output stream to make it contiguous (1 or 2 kernel options ...)
• Python code to call the OpenCL code
• Non regression test
• Profiling and performance assessment

[t20100]

I've added a commit to fix the conflict introduced by #3991.
A rebase would be best in order to sparse the merge of main in this branch.