Add mafutils info and record block aggregates in the index header

[gwct]

There was no cheap way to ask what a MAF contains. stats reads the whole file and validate only answers whether the index is trustworthy.

mafutils info prints size, compression, index freshness, block/scaffold counts and a species list. On a 42.70 GB file it takes 1.3 s (0.7 s with --sample-blocks 0), because mafutils index now records the counts in the index header and info normally reads nothing but that one line.

• New header fields (blocks, scaffolds, ref_bases, aln_cols, seq_lines, max_seqs), appended after hash=. Written identically to both index files, since validate compares the two headers for exact equality. format stays at 2.

• Three tiers, never scanning the MAF for counts: header only; else stream the index (~52 MB/s) with a warning naming its size and recommending a rebuild; else report what needs no index and point at mafutils index.

• Species are sampled from the first 1,000 blocks and labelled possibly incomplete. They aren't in the index, and collecting them during indexing measured 2.06× slower. stats remains the exhaustive source.

• Compatibility: index rows are byte-identical (verified across all 8,094,203 rows of a real index); v0.6.0 code reads a new-format index cleanly (fetch/stats/gc/validate all pass, validate VERIFIED); new code reads old indexes via the fallback. No rebuild required, it only makes info instant. Indexing cost measured at ~1.3% (4.71 s → 4.77 s on a 1.5 GB slice). A full 42.7 GB rebuild produced aggregates matching an independent awk pass exactly. 148 tests passing.

🤖 Generated with Claude Code