feat(prepro): add --retry-reverse-complement=true when running nextclade in prepro

[anna-parker]

resolves #2652

Screenshot

PR Checklist

• All necessary documentation has been adapted.
• The implemented feature is covered by appropriate, automated tests.
• Any manual testing that has been done is documented (i.e. what exactly was tested?)

🚀 Preview: https://retry-reverse-complement.loculus.org

[claude]

Claude finished @anna-parker's task in 1m 25s —— View job

---

Code Review

• Read changed files
• Analyze the change in context
• Provide review feedback

---

The PR adds --retry-reverse-complement=true to both invocations of nextclade3 run in the preprocessing pipeline — once in assign_segment_with_nextclade_align (for segment classification) and once in enrich_with_nextclade (for main alignment/QC). The flag is applied consistently across both call sites, which is correct.

What the flag does: When a sequence fails to align, nextclade retries with the reverse complement. This improves robustness for sequences submitted in the wrong orientation.

One thing worth considering: The nextclade JSON output includes an isReverseComplement field when a sequence was aligned as its reverse complement. Currently this field is not surfaced to submitters as a warning. It could be worth adding a warning annotation when result["isReverseComplement"] is True (in parse_nextclade_json or wherever the JSON metadata is processed) — so submitters know their sequence was flipped. That said, this would be a follow-up improvement and isn't a blocker for this PR.

Testing: No tests have been added or updated. Given this is a flag change to an external tool call rather than new business logic, that's reasonable — the behavior of nextclade itself can't easily be unit-tested here.

No opt-out: The flag is hardcoded to true with no config option to disable it. If there are organisms/datasets where reverse complement retry is undesirable (e.g. where strand orientation is meaningful metadata), this can't be turned off without a code change. Worth noting, but may not matter in practice.

Overall this is a sensible improvement to alignment robustness — no issues blocking merge.

[claude]

This PR may be related to: #2652 (Handling reverse complemented sequences)