Optimize: add post-download population subsetting to reduce processing overhead

[balis]

Context

The 1000 Genomes Phase 3 FTP server only provides whole-cohort VCF files (ALL.chr*.vcf.gz)
containing all 2504 samples. No population-specific files exist. This was verified by checking
every relevant directory on the server:

• /release/20130502/ — only ALL.chr*
• /supporting/bcf_files/ — only ALL.chr*
• /supporting/vcf_with_sample_level_annotation/ — only ALL.chr*
• /supporting/related_samples_vcf/ — only ALL.chr*

Problem

When an analysis targets specific populations (e.g., only EUR and AFR), the workflow still:

1. Downloads all 2504 samples (unavoidable — server limitation)
2. Processes all 2504 sample columns through individuals.py before filtering

Proposed optimization

Add a bcftools view -S <samples.txt> step immediately after data extraction (tabix/download)
to subset the VCF to only the requested populations' samples before further processing.
The panel file (integrated_call_samples_v3.20130502.ALL.panel) provides sample-to-population
mappings.

This would reduce disk usage and processing time in downstream steps (individuals.py,
mutation_overlap.py, frequency.py) proportionally to the population subset ratio.

Impact

• Network transfer: No change (unavoidable)
• Disk usage: Reduced proportionally (e.g., 2 of 5 super-populations → ~60% reduction)
• Processing time: Reduced in individuals/mutation_overlap/frequency steps

Notes

• The columns.txt balanced sampling (30 samples/population) already limits processing
  for the current test setup, but full-scale runs would benefit significantly
• The subsetting step could be added to data_preparation.steps in the plan as a
  post-extraction command