Bump quant_tximport_summarizedexperiment and quantify_rsem for deterministic collect() ordering

[pinin4fjords]

Description

Bumps quant_tximport_summarizedexperiment and quantify_rsem to pull in nf-core/modules#12377.

TXIMETA_TXIMPORT and CUSTOM_RSEMMERGECOUNTS each stage a multi-file input built by collecting per-sample process outputs on a queue channel. Queue channels emit in task-completion order, which is non-deterministic across runs, so the staged input list order - and therefore the task hash - changed run to run even when nothing else did, breaking -resume for these processes and everything downstream of them. Same pattern as the CUSTOM_TX2GENE fix already merged for this subworkflow.

Both scripts already derive sample identity from the staged file/directory names rather than input list position (TXIMETA_TXIMPORT's R template via list.files(), CUSTOM_RSEMMERGECOUNTS's script via a directory glob), so sorting the collected list by name is a no-op on outputs - it only makes the cache key reproducible.

Note the diff also carries forward the CUSTOM_TX2GENE determinism fix from #1875, since dev hasn't picked that up yet - both fixes live in the same subworkflow file.

Verification

• nf-core subworkflows lint clean for both subworkflows against the new pin.
• Upstream nf-core subworkflows test --profile docker for both subworkflows: all pass, snapshots byte-identical before/after (see Deterministic collect() ordering in tximport and RSEM merge counts modules#12377).

Closes #1879

PR checklist

• This comment contains a description of changes (with reason).
• nf-core subworkflows lint passes for both bumped subworkflows.
• CHANGELOG.md is updated.

[pinin4fjords]

Admin-merged with required-checks bypass. 3 in-flight workflow(s) cancelled: Run nf-test, Run nf-test on ARM, Run GPU nf-tests.

Rationale: the unverified diff since the last fully-green commit (1889d16b5) was just the two git_sha bumps in modules.json pulling in nf-core/modules#12378. Every other standard workflow (Run nf-test, Run nf-test on ARM, Run GPU nf-tests, Nextflow strict syntax lint, nf-core template version comment) was already green at that commit - the only failure was nf-core linting's check_local_copy check, which is exactly what this bump fixes. modules.json isn't read by Nextflow at runtime, so it can't affect the nf-test/GPU suites already verified green. Locally re-ran nf-core subworkflows lint for both bumped subworkflows against the new pin (clean), and by the time the merge went through, the nf-core linting run itself had actually finished on its own and come back green too.