Refactor how prepro enrich with nextclade function handles multiple segments

[anna-parker]

raised by @corneliusroemer during code review of #5382 (reference)

enrich_with_nextclade modifies dictionaries in place - we should refactor this to have clear input and output and less pre-initalization of dictionaries be required

[theosanderson-agent]

This appears to have been substantially addressed by PR #5554 ("deduplicate segment assignment code and add tests", which explicitly says "partially resolves #5576") combined with PR #5800 ("multi-pathogen refactor"), which is a major restructuring with improved typing, a new SequenceName type, and restructured config handling. Together these PRs address the core concerns about dictionary mutation and unclear I/O in the nextclade enrichment function.