RaPDTool v2.3.1

A maintenance release: three fixes to input handling and report formatting, and a
measured change to the FOCUS reporting cutoff. The pipeline, its modes and its outputs
are otherwise unchanged from v2.3.0.

Fixed

• The documented input extensions now actually work. FOCUS selects its inputs by file
  suffix and reads only .fna, .fasta and .fastq. Inputs documented as valid — .fa,
  .fas, .ffn, .frn, .fq, and anything .gz — never reached the profiling step, and
  the failure surfaced as a Python traceback rather than a message. The working copy is now
  decompressed if needed and renamed to a suffix FOCUS accepts. Output directories keep the
  name of the file supplied.

• Long names no longer wrap in rapdtool_confidence.tbl. Table columns are sized to
  their widest value instead of sharing a fixed width budget, so a long species name or
  taxID stays on one line. Only the scaffold list is narrowed, and it remains complete in
  rapdtool_confidence.txt.

Changed

• The FOCUS reporting cutoff drops from 1 % to 0.5 % relative abundance. The operating
  point comes from the threshold sweep of the RaPDTool benchmark: across mock communities
  at 3, 10 and 30 M reads, 0.5 % recovers every species present (recall 1.0) against
  0.80–0.95 at 1 %, with a higher F1 at every depth tested. This affects only which rows
  are listed in rapdtool_confidence.tbl|txt — the complete FOCUS profile under
  profilesfmbm/ is unfiltered as before, and Mash-based detection is unaffected.

• The report merger now uses Text::SimpleTable directly instead of
  Text::SimpleTable::AutoWidth, which also drops the Moo and Type::Tiny dependencies
  from the image.

Documentation

• Screen mode takes a single file in -i, so paired or multi-lane FASTQ must be
  concatenated first (cat R1.fastq R2.fastq > all.fastq). Noted in the README, the usage
  block and the launcher help.

Install / upgrade

conda create -n rapdtool -c conda-forge -c kjestradag rapdtool
# already installed:
conda clean --index-cache && conda update -n rapdtool -c conda-forge -c kjestradag rapdtool

The Apptainer image and the reference databases are downloaded and cached on first use;
an existing cache refreshes itself when a newer image is published, so no manual step is
needed.