fix: prevent storeDir race condition with parallel chromosome tasks on cloud executors

[ikhrustalev]

Summary

Fixes #481

When running multiple chromosomes in parallel on cloud executors (e.g. Google Cloud Batch with GCS FUSE), processes using storeDir write versions.yml to the same path concurrently. The data files don't collide (they include chromosome in the filename), but versions.yml is always the same name, causing stale file handles or missing output files.

This PR partitions storeDir by sample ID and chromosome using closures for lazy evaluation, so each task gets its own isolated directory while preserving cross-run caching.

Before:

cachedir = params.genotypes_cache ? file(params.genotypes_cache) : workDir
storeDir cachedir / "genomes" / "recoded"

After:

storeDir { (params.genotypes_cache ? file(params.genotypes_cache) : workDir).resolve("genomes/${meta.id}/recoded/${meta.chrom}") }

Changes

• 14 files changed, all modules using storeDir
• Replaced eager path construction with closure-based lazy evaluation
• Added meta.id and meta.chrom (where available) to storeDir paths

Testing

• Standard test profile passes locally (nextflow run . -profile test,docker)
• Verified on Google Cloud Batch with 22-chromosome VCF input (sample HG00096, PGS002209 with --run_ancestry)
• Pipeline completes with zero errors vs. 17+ retry errors before the fix
• Scores are identical to runs using the original code