Staden io_lib 1.14.10

WARNING: some bugs have been found in 1.14.10. Use for evaluation of CRAM 4 only, while we track down these. A 1.14.11 will be available once we've fixed the problems. Apologies for the inconvenience.

Updates

• BAM: Libdeflate support (https://github.com/ebiggers/libdeflate).
  This library is significantly faster than zlib, so it is a good
  alternative to the Cloudflare and/or Intel libraries.

  Configure using --with-libdeflate=/dir/to/deflate/install

• CRAM EXPERIMENTAL: Added custom quality and identifier codecs.
  Also added the ability to use libbsc as a general purpose codec.

  These are NOT OFFICIAL and so not enabled by default (version 3.0).
  However as a technology demonstration only, they are available with
  scramble -V3.1 or -V4.0 for evaluation and to promote discussion on
  future CRAM formats. Do not use these on production data.

  Implementations of the codecs and CRAM version 4.0 layout are liable
  to change without prior warning.

• CRAM: name sorted files now automatically switch to non-ref mode.

Bug fixes

• CRAM: Considerable fixes to multi-threading.

  • Using more than 1 slice per container with threading now works.
  • Removal of race conditions when using CRAM_OPT_REQUIRED_FIELDS.
  • Combinations of ref and no-ref mode in adjacent containers.
  • Other misc. threading bugs.

• Corrected end-of-range check in some scenarios.

• CRAM: bug fix to index creation when a slice contains exactly one
  alignment.

• SAM: fixed parsing of illegal sequence characters (eg "Z").
  These are now treated as "N" and not "=".

• BAM/SAM: protect against out of bound CIGAR operations.

• CRAM: hardening of rANS codec against malicious input.
  Also fixed a very rare frequency renormalisation case.

• CRAM: fix with range queries used in conjuction with turning off
  sequence retrieval (via CRAM_OPT_REQUIRED_FIELDS).

• Improved test harness for Windows and some header file problems.

• Fixed bgzip on big endian systems. (Debian bugs 876839, 876840)