feat(rag): switch embedding model to BioLORD-2023-C

Replace all-MiniLM-L6-v2 with FremyCompany/BioLORD-2023-C.

BioLORD is anchored to UMLS/SNOMED CT/MeSH ontologies during training,
so it natively resolves biomedical synonyms (TARDBP↔TDP-43, SOD1↔superoxide
dismutase) without relying on exact string matches. This improves retrieval
quality for physician queries that use clinical phrasing rather than gene
symbol abbreviations — the primary trust signal for a physician-facing tool.

ChromaDB collection rebuilt: 510 chunks at 768d (vs 384d for MiniLM).
Smoke test: query 'superoxide dismutase mutation motor neuron' returns
all SOD1-specific papers at top-3 with no abbreviation in the query.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

rag/indexer.py

@@ -13,7 +13,9 @@ from models import ALSPaper
 
 _logger = get_logger("rag.indexer")
 
-_EMBED_FN = SentenceTransformerEmbeddingFunction(model_name="all-MiniLM-L6-v2")
+# BioLORD-2023-C: anchored to UMLS/SNOMED CT/MeSH ontologies — natively understands
+# biomedical synonyms (TARDBP = TDP-43, SOD1 = superoxide dismutase) and clinical phrasing.
+_EMBED_FN = SentenceTransformerEmbeddingFunction(model_name="FremyCompany/BioLORD-2023-C")
 
 
 def _chunk_paper(paper: ALSPaper) -> list[dict]: