fix: combine all membrane markers for cellSAM segmentation

[mikemcka]

Problem

When running cellSAM whole-cell segmentation, only the first matching
membrane marker was ever used. extract_channels in bin/cellsam_segment.py
loops over the --membrane-channel list and breaks on the first hit, so any
additional markers were silently discarded. With a single weak boundary marker,
cellSAM falls back to a nucleus-driven (round) segmentation.

This is inconsistent with the Cellpose/SOPA path, which already runs
COMBINECHANNELS to max-combine all listed membrane markers into one channel
before segmentation.

Fix

Wire COMBINECHANNELS into the CELLSAM_SEGMENT subworkflow (mirroring
subworkflows/local/sopa_segment/main.nf): all listed membrane markers are
max-combined into a single combined_membrane channel, and that combined tiff
is what feeds both CELLSAMWC and CELLSAMNUC. The duplicate COMBINECHANNELS
call in the report block is removed (it now runs upstream).

CELLSAM_SEGMENT_WBACKSUB inherits the fix automatically — it background-subtracts
then delegates to CELLSAM_SEGMENT.

Observability

Because cellSAM now only sees the single combined channel (its
Available channels: log no longer shows the source panel), combine_channels.py
logs what it combined to stderr, e.g.:

[combine_channels] nuclear='DAPI'; membrane=['CD45', 'Ecad'] -> 'combined_membrane' (method=max, n=2)

This benefits the Cellpose path and the report background too, since they share the module.

Tests

• New combinechannels - multi-membrane combine case (the module previously had
  no multi-membrane coverage — both existing cases use a single channel).
• New cellsam_segment - multi-membrane combine case guarding the wiring.
• Both reuse tests/data/comet/synthetic_multichannel.tif (CD45:TRITC = two markers);
  no new test binary.

Validation already run

• Ran the real combine_channels.py on a synthetic 2-marker array (markers in
  non-overlapping halves): output is [DAPI, combined_membrane] with
  combined_membrane == max(CD45, Ecad) and not equal to either single marker.
• Full Nextflow -stub-run of the cellSAM path with CD45:Ecad:
  COMBINECHANNELS -> CELLSAMWC/CELLSAMNUC -> CELLMEASUREMENT -> SEGMENTATIONREPORT,
  pipeline completed successfully (no double-COMBINECHANNELS error).

Reviewer checklist — snapshots need blessing on a GPU node

nf-test snapshots are not included (no nf-test/GPU on the dev box). Before merge, run with --update-snapshot and commit the .snap files:

• modules/local/combinechannels — new case (deterministic, GPU-free)
• subworkflows/local/cellsam_segment — new case and re-bless the existing case (changed: COMBINECHANNELS now runs in the workflow → new entry in versions)
• subworkflows/local/cellsam_segment_wbacksub — re-bless the existing case (same reason)

🤖 Generated with Claude Code

[github-actions]

This PR is against the main branch ❌

• Do not close this PR
• Click Edit and change the base to dev
• This CI test will remain failed until you push a new commit

---

Hi @mikemcka,

It looks like this pull-request is has been made against the mikemcka/sp_segment main branch.
The main branch on nf-core repositories should always contain code from the latest release.
Because of this, PRs to main are only allowed if they come from the mikemcka/sp_segment dev branch.

You do not need to close this PR, you can change the target branch to dev by clicking the "Edit" button at the top of this page.
Note that even after this, the test will continue to show as failing until you push a new commit.

Thanks again for your contribution!

[mcmero]

Looks good, I'll merge it to dev.