Release v2.1.0

• Added compilation to standalone C files.
  -C compiles to stdout,
  -o gives an output filename,
  -p gives a C function prefix (default: cmph_c_)
  It's usually about 10% faster than loading the serialized .mph file at run-time.
• Added ordering_table support needed for randomnly ordered algos (all but CHM):
  -r creates a compressed table to lookup the real key indices
  -R creates a new ordered keys file, so that cmph_search in that new keys file
  will return the proper index.
• Optimize bmz, chm search with hash_vector (getting multiple hash values per call)
• When selecting one hash function with -f, it selects this hash function
  for the next hash functions also, to override the default jenkins.
• Extended bm_numbers to accept -h hash, -a algo, -s size, -i iters.
• Added bm_compiled benchmarks.
• mphf_tests got -s, -a and -f support.
• Added -Wextra
• Removed unused args in the internal API.
• Fixed a lot of memory leaks.

Full Changelog: https://github.com/rurban/cmph/commits/2.1.0

Unofficial v2.0.3

Fixed hash states, re-enabled the other hashes, added wyhash and HW crc32.
This fixed bzm8, chd and chd_pc.
Minor fixes for buffer overflows, leaks and unaligned accesses.
Fixes for make distcheck.
Proper docs generation.