Add cross-encoder rerank button + dual embedding models (general/biomedical) with live switch

Plot: tier marker size + line width per method so overlapping hits nest as concentric rings

Refit UMAP on Python 3.10 to match Space runtime (numba pickle compat)

UMAP projection + filtered plot + expandable abstracts + cardiovascular example queries

Add retrieval comparison demo (BM25/Dense/Hybrid) + precomputed artifacts

initial commit