V0.0.7

[alejandrogzi]

Added

• New bed subcommand — chaintools bed converts chain files to BED format,
  supporting BED3, BED6, and BED12 output. Chains can be read from standard input,
  explicit file arguments (-c/--chains), or path-list files (-f/--file).
  Options include --spanned (emit the full chain span as a single block),
  --sort-by (coordinate or score ordering), --side (reference or query
  coordinates), --type (BED variant selection), --out-bed (write to file),
  and --gzip (compress output). Input/output path conflict detection prevents
  accidental overwrites.

• Chain construction from absolute blocks — OwnedChain::from_absolute_blocks
  builds complete owned chains from absolute alignment coordinates, converting
  them to UCSC dense chain blocks via absolute_to_dense_blocks. This enables
  callers to accumulate aligned intervals first and derive block gaps later, which
  is the foundation for axtChain-compatible workflows.

• AbsoluteBlock model type — A new coordinate representation for alignment
  blocks using absolute reference/query start and end positions, with validation,
  interval-length helpers, and overlap detection. The absolute_to_dense_blocks
  function converts a sorted, non-overlapping list into the UCSC dense block
  format (computing gap sizes from neighboring absolute coordinates).

• Metadata writer API — write_metadata_line and write_metadata_lines in
  the io::writer module allow writing comment and metadata lines to chain
  output streams. These are now unconditionally exported (previously gated behind
  the sequence feature).

• Public scoring functions — score_ungapped_slices, score_absolute_block,
  and score_absolute_blocks are now exported from the seq::score::chainscore
  module, enabling UCSC chainScore computation directly from absolute alignment
  coordinates and ungapped slices.

• SequenceResolver convenience methods — preload (alias for
  new_filtered), sequence (borrow a full sequence slice), and sequence_len
  (query decoded sequence length) make the resolver easier to use without
  allocating copied ranges.

• Standalone revcomp module — Reverse-complement logic extracted from
  antirepeat into seq::revcomp, with complement_base,
  reverse_complement_in_place, and reverse_complement. Full IUPAC ambiguity
  code support and case preservation, with dedicated unit tests.

Changed

• write_chain_dense, write_chain_header, and write_dense_blocks are no
  longer behind the sequence feature gate; they are always available.
• Import ordering and minor formatting alignment across merge, sort, and
  split modules.
• The --level short flag -L was removed from the global CLI to avoid
  conflicts with subcommand flags; --level (long form) remains.

Documentation

• bed subcommand is integrated into the CLI help and dispatch.

Notes

• 94 binary tests + 83 lib/integration tests pass, including 18 new ones
  spanning BED output correctness, chain-from-absolute-blocks construction,
  absolute block validation and gap computation, metadata writer safety,
  revcomplement IUPAC handling, sequence resolver preloading, and scoring of
  absolute blocks with gap cost. gzip-feature tests pass too.
• End-to-end: confirmed chaintools bed produces valid BED12 records on
  real chain files, with correct coordinate ordering and block structure.