v1.6.0 — uniqsketch homopolymer filter and strict margin

Features

• uniqsketch --max-homopolymer=N: drop signatures containing long homopolymer runs.
  Signature candidates with a single-base run longer than N bp are rejected. The aggregate entropy
  filter scores a signature across the whole k-mer, so a short local run (a poly-G tract, for example)
  gets diluted and can survive; this adds a direct local-run-length gate for such indel-prone tracts.
  The check runs while each genomic selection slot is scanned, so a rejected candidate is replaced from
  the same slot wherever one exists and per-reference signature counts and genomic spacing are largely
  preserved. The default (--max-homopolymer=0) disables the filter and leaves selection unchanged.
• uniqsketch --strict-margin: enforce --min-margin as a hard requirement.
  By default --min-margin=2 falls back to a margin-1 candidate when a selection slot has no 2-safe
  option. With --strict-margin such slots are left empty instead, so every emitted signature satisfies
  the requested margin, at the cost of a smaller sketch. The default (off) leaves --min-margin
  behavior unchanged.

Both options are off by default, so this release is backward compatible with v1.5.0 output.

Verification

Measured on an 8-genome simulated reference set:

• Without the homopolymer filter, 61 of 799 signatures carry runs longer than 5 bp (longest 8). With
  --max-homopolymer=5, none exceed the bound (longest exactly 5) and only 26 signatures are lost net,
  because slot selection substitutes another candidate.
• With --min-margin=2, soft mode yields 800 signatures of which 737 (92.1%) satisfy margin ≥ 2;
  --strict-margin yields 737 signatures, all of which satisfy it.
• An end-to-end query against a homopolymer-filtered index recovered a 90/10 two-reference mixture at
  90.03% / 9.97%, so quantification accuracy is preserved.

Credits

The --max-homopolymer and --strict-margin features were contributed by Christian Gauthier (@chg60).

Full changelog: https://github.com/amazon-science/uniqsketch/blob/main/CHANGELOG.md