Suppress repetitive GeneMark license banner in training/predict logs

[nextgenusfs]

Problem

During funannotate2 train (and predict), gmhmme3 is invoked once per contig. Each invocation prints a two-line license banner to both stdout and stderr:

GeneMark.hmm eukaryotic 3
16 days remaining in the license period.

runSubprocess captures both streams and forwards them to logger.debug(...), and the file handler is set to DEBUG. Result: the log file gets flooded with duplicate banners (one block per contig — typically dozens to thousands).

Example from a user's log:

[Apr 22 08:53 AM] gmhmme3 -f gtf -k 0.03 -m gmhmm.mod -o g_272.fa.gtf g_272.fa
[Apr 22 08:53 AM] GeneMark.hmm eukaryotic 3
16 days remaining in the license period.

GeneMark.hmm eukaryotic 3
16 days remaining in the license period.

[Apr 22 08:53 AM] gmhmme3 -f gtf -k 0.03 -m gmhmm.mod -o g_947.fa.gtf g_947.fa
...

Fix

1. funannotate2/utilities.py — add an opt-in output_filter=None kwarg to runSubprocess. When set, the success path delegates output handling to output_filter(logfile, stdout, stderr) instead of the default logoutput(logfile.debug, process). Default behavior is unchanged for every existing caller. Error path (CalledProcessError) is untouched.

2. funannotate2/abinitio.py — add a module-level _GENEMARK_BANNER_LOGGED flag and a _genemark_output_filter helper. The helper splits each stream into banner lines (matching GeneMark.hmm eukaryotic or days remaining in the license period) vs. other content; the banner is logged at debug once per process, then suppressed on subsequent calls. Non-banner content (real warnings/errors) continues to go through log.debug. Both gmhmme3 call sites (run_genemark and the evaluation loop's tool == "genemark" branch) now pass output_filter=_genemark_output_filter.

Preserved behavior

• The per-contig gmhmme3 -f gtf -k 0.03 ... command line is still logged (that one is useful for debugging; only the banner is suppressed).
• Any non-banner gmhmme3 output (real warnings/errors) is still logged at debug.
• Command failures (CalledProcessError) still log full stdout/stderr at error level — unchanged.
• No behavioral change for Augustus / SNAP / GlimmerHMM / BUSCO or any other tool.

Verification

• python3 -m py_compile funannotate2/utilities.py funannotate2/abinitio.py — both compile clean.
• Ad-hoc behavior test against a stub logger:
  • First call with banner content → 1 debug emission, flag flips to True.
  • Second identical call → still 1 (banner suppressed).
  • Third call with banner + extra content → 3 (banner still not re-emitted, non-banner lines forwarded).
  • Empty / None inputs handled without raising.

Risk notes

• Module-level flag lives for the Python process. Current call sites are serial per-contig in the main process. If a future caller moves gmhmme3 into a multiprocessing worker (fork), each worker would log the banner once — still a massive reduction.
• Banner detection is substring-based. If GeneMark ever changes the banner wording significantly, a new line would fall through to normal debug logging (degrades to current behavior — non-fatal).

---

Pull Request opened by Augment Code with guidance from the PR author