fix(prepro): Allow nextclade metadata fields to be per_segment

[anna-parker]

preview URL: https://prepro-fix.loculus.org/

Fixes

In #2100 I mentioned that there is a bug in the prepro code when processing multi-segment nextclade alignments.

When all segments are merged:

When the L segment is uploaded individually:

Basically these alignment metrics need to be annotated as per_segment.

Summary

• The PR allows us to have metadata fields that were created by nextclade also per_segment. In order to do this the per_segment parameter must also be added as a field to the preprocessing arguments. For example:

- name: total_unknown_nucs
        type: int
        header: "Alignment states and QC metrics"
        noInput: true
        rangeSearch: true
        per_segment: true
        preprocessing:
          args: { type: int, per_segment: true }
          inputs: { input: nextclade.totalMissing }

• I then adjusted the prepro code to use this argument when updating the metadata fields.
• This PR also cleans up the prepro file a bit by moving some functions to another script as prepro is getting a bit too large.

Screenshot

[corneliusroemer]

This looks great! Thanks for tackling!

I've opened an issue #2115 to improve how we show per-segment metadata in a more structure way.

I've also formatted with ruff (it didn't seem to have been) and fixed mypy errors. These might have already pre-existed from the cchf merge (I haven't yet reviewed the prepro part of the code).

See commit 8d4cb27 for my changes

[corneliusroemer]

I just realized this PR makes the metadata spec very undry - I don't think this is necessary:

The PR allows us to have metadata fields that were created by nextclade also per_segment. In order to do this the per_segment parameter must also be added as a field to the preprocessing arguments.

[corneliusroemer]

Simplification attempt: https://github.com/loculus-project/loculus/pull/2131/files