Fix input_maf validation, publishDir glob, cache, schema, and add MAF input docs

[Copilot]

Review feedback on the MAF input feature revealed several correctness issues: invalid publishDir globs silently dropping VCFs, disabled cache false on a non-deterministic process, misleading schema description, and no early validation when --input_maf is used without --use_custom_depths.

Fixes

• conf/modules.config: "**{vcf}" → "*.vcf" for MAF2VCF and INPUTMAF2VCF — the original pattern is not a valid glob and published nothing
• modules/local/downsample/mutations/main.nf: Restore cache false — downsampling is non-deterministic; cached results across reruns are incorrect by definition
• workflows/deepcsa.nf: Add early guard that calls error when --input_maf is set without --use_custom_depths true, preventing silent fallback to INPUT_CHECK

if ( params.input_maf && !params.use_custom_depths ) {
    error "ERROR: '--input_maf' requires '--use_custom_depths true' and a matching '--custom_depths_table'. ..."
}

• nextflow_schema.json: Rewrote input_maf description/help_text — removes the misleading "WILL OVERWRITE THE input.csv DATA" wording; accurately states the use_custom_depths dependency and that --input is still required

Tests

• Added tests/test_data/test_mutations.maf (minimal non-deepCSA MAF)
• Added nf-test case "Should fail when --input_maf is provided without --use_custom_depths" asserting workflow.failed and the expected error string
• Added commented-out integration test template for the full input_maf + use_custom_depths path (requires a precomputed depths table as test asset)

Documentation

Added docs/usage.md section "MAF file as input (alternative input mode)" covering: when to use it, required parameter combination, MAF column requirements (deepCSA-origin vs. external), example invocation, and pipeline behaviour.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.