v1.1.0

Added

• --miller-madow option for regional-mds (and miller_madow=True on
  EndMotifsIntervals/BreakpointMotifsIntervals's motif_diversity_score/
  mds_bed) applies a Miller-Madow bias correction to rMDS, counteracting
  the downward bias of the plug-in Shannon entropy estimator in regions
  with few fragments. Off by default; existing rMDS output is unchanged.

Changed

• The package version is now derived from git tags via setuptools-scm
  instead of a hand-edited version string: 1.0.0 exactly on a tag,
  1.0.1.dev6 six commits past one. No action needed for normal installs;
  contributors on an editable install (pip install -e .) should re-run
  that command after pulling new commits if they want finaletoolkit --version to reflect the latest commit.

Fixed

• filter-file/filter_file now accepts a -r/--reference (reference_file)
  option. Previously CRAM input to filter-file always failed with "CRAM
  files require a reference file", since the CLI's CRAM validation is shared
  across every subcommand but filter-file had no way to satisfy it. CRAM
  input is converted internally to BAM for processing and the filtered
  output is written back out as CRAM (using the given reference) when the
  input was CRAM.
• cleavage_profile/multi_cleavage_profile no longer risk an out-of-memory
  crash on large intervals. The per-position depth/end-count calculation
  built an (n_fragments, n_positions) boolean matrix, which could reach
  hundreds of gigabytes for a single interval (e.g. tiled BED bins that get
  merged into one large region); it's now computed with an O(positions)
  coverage-difference array instead, with identical output.