CSV and JSON search outputs have different shapes #19

The CSV branch of /api/v1/search writes raw DB rows, while JSON maps them through CLEANDataBase. As a result:

• Column names differ: protein_name/protein, ncbi_taxid/ncbi_tax_id, enzyme_function/function, protein_sequence/sequence, uniprot_id/uniprot.
• CLEAN predictions are two parallel ;-joined columns in CSV but predicted_ec: [{ec_number, score}] in JSON.
• A column change has to be made in two places.

Aligning them (flattening the model for CSV, or array_to_string in SQL) would change existing CSV column names, so it needs a decision about CSV consumers.

Raised in review of #17.