Refine mean/median sequence length output

[ewels]

Builds on #203 / 54b336e (mean & median sequence length). Two small refinements plus the test snapshots that commit didn't include.

Changes

1. Mean Length → 1 decimal place (BasicStats.java)

Currently the mean is integer division (totalBases/actualCount), so a 150.7 bp mean prints as 150. This reports it to 1 d.p. instead:

if (actualCount == 0) return "0.0";
return String.format(Locale.ROOT, "%.1f", (double)totalBases/actualCount);

(Locale.ROOT keeps the decimal separator as . regardless of system locale; the zero guard avoids an integer divide-by-zero on empty input.)

2. Median tie-breaking (SequenceLengthDistribution.medianLength())

For an even number of reads, the current code returns the upper of the two central values. This uses the standard definition — the mean of the two central values, rounded up:

• Odd N: the single central value.
• Even N: (lo + hi + 1) / 2.

Also added a total == 0 guard (returns 0) so an empty distribution doesn't throw. Identical to current behaviour for uniform-length (short-read) data; differs only on even read counts with two distinct central lengths.

3. Test snapshots (FileContentsTest_{minimal,complex}_fastqc_*)

Updated the approved data + HTML files for the new Mean Length / Median Length rows. The HTML snapshots are edited in place so the embedded base64 chart images are left untouched (rendering differs per OS).

Notes

• Labels (Mean Length / Median Length) and filtered-sequence handling are left exactly as in 54b336e.
• Heads-up: the approved fastqc_data.txt files still carry ##FastQC 0.12.2.devel while the build now emits 0.13.0.devel — a pre-existing version-string mismatch in the snapshots, independent of this change. Left as-is to keep this PR scoped; happy to bump it if you'd prefer.

🤖 Generated with Claude Code