fix: count secondary+supplementary records as secondary only (flagstat)

[BenjaminDEMAILLE]

Fixes #125.

Problem

Records carrying both SECONDARY (0x100) and SUPPLEMENTARY (0x800) were counted in both the secondary and supplementary flagstat totals, because the two bits were tested independently:

if is_secondary { self.secondary += 1; }
if is_supplementary { self.supplementary += 1; }

samtools gives SECONDARY priority — bam_stat.c (flagstat) and stats.c both classify such a record as secondary and never reach the supplementary counter. The double count inflates the supplementary total and breaks the invariant primary + secondary + supplementary == total.

While checking this against samtools, a second instance of the same conflation showed up in the samtools stats SN section: non-primary alignments is nreads_secondary in samtools, but RustQC emitted secondary + supplementary. That line was wrong for any BAM containing supplementary alignments, dual-flagged or not.

Fix

• accumulators.rs: if is_secondary { .. } else if is_supplementary { .. }
• stats.rs: non-primary alignments: now reports secondary alone

Verification

Ground truth from samtools 1.24 on a synthetic BAM (4 primary pairs, 2 secondary-only, 3 supplementary-only, 5 dual-flagged):

samtools flagstat            samtools stats
18 + 0 in total              SN  non-primary alignments:  7
 8 + 0 primary               SN  supplementary alignments: 3
 7 + 0 secondary
 3 + 0 supplementary

A second control file with the dual-flagged records removed (2 secondary-only, 3 supplementary-only) gives non-primary alignments: 2, confirming that line tracks secondary alone rather than the sum.

New unit test test_dual_flagged_reads_count_as_secondary_only builds the same record mix and asserts those numbers plus the sum invariant.

• cargo test — 201 lib + 18 integration tests pass
• cargo fmt --check, cargo clippy -- -D warnings clean

🤖 Generated with Claude Code