feat: implement hallucination detection

[DelusionalSimon]

Add an internal verification step to cross-check the LLM's claims (gene names, locus tags) against the input annotation file.

Tasks

• Create a new helper function that runs after the LLM call.
• This function will parse the potential_features and check if the supporting_genes and locus_tags exist in the original data.
• Define the additions to the JSON scheme (how explicit to be in the tag)
• The function will add a new key to each feature in the JSON, ("verification_status").
• If a gene or locus tag is not found or gets mismatched, it will tag it appropriately ("verification_status": "possible hallucination").
• The helper function for the human-readable .txt report will add some notes on detected hallucinations in the output
• Test thoroughly with known-good and known-bad data.
• Update documentation to explain the verification status tags.
• (Optional) look into more advanced hallucination detection techniques (RAG)