Add optional STAR alignment with isolated BAM and junction outputs

[Fuki-Kudoh]

Supersedes #15 and implements issue #14.

Motivation

Add optional STAR genome alignment while preserving the existing
Salmon/kallisto quantification workflow and output contract.

Implementation

• Added --aligner {none,star} with default none.
• Added validation and provenance tracking for STAR indices.
• Added a conditional align phase using trimmed FASTQs.
• Added paired-end and single-end STAR support.
• Added two-pass STAR alignment with coordinate-sorted BAM output.
• Added BAM indexing with samtools.
• Stored STAR outputs under align/star/<sample>/.
• Preserved existing Salmon/kallisto outputs under abundance/.
• Added resume handling and STAR index change detection.
• Added unit and fake-tool integration tests.
• Updated CI, README, and software specification.

Outputs

Per sample:

• Aligned.sortedByCoord.out.bam
• Aligned.sortedByCoord.out.bam.bai
• SJ.out.tab
• ReadsPerGene.out.tab
• Log.final.out

Human real-data validation

Validated on a 500,000-pair subset of GSE157878 BMDM RNA-seq.

• Reference: mm10
• Layout: paired-end
• Strandedness: none
• Threads: 12
• Pipeline completed successfully.
• BAM passed samtools quickcheck.
• Unique mapping rate: 68.51%.
• Reads mapped to multiple loci: 9.73%.
• Total detected splices: 333,024.
• STAR and Salmon outputs were both generated.
• A second identical invocation skipped every completed phase,
  confirming resume behavior.

[Fuki-Kudoh]

Human real-data validation

Validated manually by Fuki Kudoh.

Dataset

• GEO series: GSE157878
• Sample: BMDM
• Input: deterministic 500,000-pair subset
• Layout: paired-end
• Reference: mm10
• Strandedness: none
• Threads: 12
• WulfRNA commit: f8cd86a

Results

• The complete WulfRNA pipeline finished successfully.
• STAR and Salmon outputs were generated in isolated output paths.
• samtools quickcheck completed with exit code 0.
• STAR input reads: 500,000.
• Uniquely mapped reads: 68.51%.
• Reads mapped to multiple loci: 9.73%.
• Total splices: 333,024.
• BAM, BAI, SJ.out.tab, ReadsPerGene.out.tab, Log.final.out,
  quant.sf, gene expected-count matrix, and gene TPM matrix were
  non-empty.
• A second identical run skipped all completed phases:
  fastqc_raw, cutadapt, fastqc_trimmed, align, quant, aggregate,
  and multiqc.

Interpretation

This validates real paired-end execution, BAM integrity, splice-junction
output, coexistence with Salmon quantification, and phase-level resume.
It is an execution validation rather than a biological benchmark of
STAR counts against Salmon estimates.