v0.2.0

Added

• compress --estimate — predict the compression ratio and archive size
  for a FASTQ without writing an archive. --estimate tsv emits machine-readable
  output, and info/verify now accept multiple files or directories for batch
  reporting.
• Reference sequence coder for --max/reorder — a SPRING-style codec that
  2-bit-packs the assembled global reference and entropy-codes it with a
  clean-room LZMA, adopted only when it beats the raw representation.
• --order shuffle renumber mode — a true SPRING-style read renumbering that
  discards the input permutation, reaching SPRING-competitive ratios on datasets
  where read order carries no information.
• CLI run feedback — a TTY-aware progress spinner and human-readable
  compress/decompress run summaries (suppressed under --quiet and when stderr
  is not a terminal).

Changed

• Quality coding — fqzcomp now models quality over the symbol alphabet that
  actually occurs in the file rather than a fixed 0..QMAX range, shrinking the
  quality stream on data with sparse quality alphabets. Byte-identical
  round-trips are preserved.

Performance

• Reorder merge — the overlap-merge k-mer index now uses a rolling hash with
  sharding (~10% faster --max, byte-identical output), and the
  merge_reference vote-scatter is parallelized.

Documentation

• VHS-rendered terminal demos on the docs site, refreshed benchmarks (including
  an fqxv-shuffle row), and clearer positioning.