feat: módulo embedding — encode, model CRUD, stores sqlvec y usearch

Funciones Python para embeddings: carga/guardado de modelos, encoding de
texto, y almacenamiento/búsqueda vectorial con sqlite-vec y usearch.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

python/functions/embedding/embedding_search_usearch.md

@@ -0,0 +1,37 @@
+---
+name: embedding_search_usearch
+kind: function
+lang: py
+domain: infra
+version: "1.0.0"
+purity: impure
+signature: "def embedding_search_usearch(path: str, query_embedding: list, k: int = 10, dim: int = 384) -> list"
+description: "Busca los k vecinos mas cercanos en indice USearch persistido. Busqueda sub-milisegundo."
+tags: [embedding, usearch, vector, search, retrieval, ann, python]
+uses_functions: []
+uses_types: []
+returns: []
+returns_optional: false
+error_type: "error_go_core"
+imports: [usearch, numpy]
+tested: false
+tests: []
+test_file_path: ""
+file_path: "python/functions/embedding/usearch_store.py"
+---
+
+## Ejemplo
+
+```python
+model = embedding_load_model(".local/models/e5-small")
+q_emb = embedding_encode(model, ["¿Que es machine learning?"], mode="query")[0]
+
+results = embedding_search_usearch("docs.usearch", q_emb, k=5)
+# [{"key": 0, "distance": 0.82}, {"key": 1, "distance": 0.65}, ...]
+```
+
+## Notas
+
+Carga el indice completo a RAM antes de buscar. Cold start ~190ms para 50k vectores.
+Busqueda aproximada (HNSW) — puede no encontrar el vecino exacto pero es 150x mas rapido que brute-force.
+Distance es inner product (mayor = mas similar, al reves que sqlite-vec).