Bzip3

* increase the maximum allowed amount of parallel workers to 64.
* clean up the cm code
* set up pkg-config support
* CLI robustness improvements

* fix a serious stdin/stdout CRLF bug on Windows that corrupted the data.
* imply `-c` when a stream isn't opened thus preventing potential UB.
* bzip3 file format documentation.
* increase the maximum amount of parallel workers to 24.
* prevent accidentally overwriting output; add the `-f` command-line flag.

* fix memory UB in libsais
* restructure src/cm.c
* block size check in src/libbz3.c
* fix shift UB in lzp
* -h/-v CLI switches
* change maximum block size to ~512M