feat: mean and median sequence length in Basic Statistics

[ewels]

Implements s-andrews/FastQC#203: report mean and median read length in the Basic Statistics module, so downstream tools (e.g. MultiQC) no longer have to estimate them from the binned length distribution.

What changed

Two new rows in the Basic Statistics table, right after Sequence length:

Sequence length          16
Mean sequence length     16.00
Median sequence length   16

• Mean = total bases / total sequences, to 2 decimal places.
• Median = derived from a per-length histogram over non-filtered sequences. For an even number of reads, the two central values are averaged and rounded up.

Both values flow through write_text_report, so they appear in fastqc_data.txt and the HTML report (default and --template modern), which render the table from the same source.

Notes / decisions

• Computed only over non-filtered sequences, consistent with the existing min/max length range.
• Histogram is kept locally in BasicStats rather than shared with SequenceLengthDistribution, which bins lengths and includes filtered reads — different semantics. Memory cost mirrors the histogram that module already maintains.
• This is a deliberate divergence from the Java text output, which only reports the range.

CI / tests

• Unit tests and the cargo test approved-output tests are updated and passing.
• ⚠️ The equivalence job (Rust vs. stored Java reference) is expected to fail for now, since the Java reference doesn't yet emit these rows. Per plan, this PR stays open until the feature lands in upstream Java FastQC; once reference data is regenerated from a build that includes it, equivalence goes green. No equivalence patches added on purpose.

🤖 Generated with Claude Code

[ewels]

On hold until upstream merge is complete.