feat: add full_text_search() to CoordinodeClient and LangChain keyword_search()

[polaz]

Feature

Add full-text search capability to the Python SDK once CoordiNode DB implements full-text search procedures.

API design

coordinode core client:

results = client.full_text_search(
    label="Entity",
    property="name",
    query="machine learning",
    top_k=10,
)
# → List[NodeResult]

langchain-coordinode adapter (CoordinodeGraph):

results = graph.keyword_search(query="machine learning", label="Entity", top_k=10)
# → List[Dict[str, Any]]

llama-index-coordinode adapter (CoordinodePropertyGraphStore):
The LlamaIndex PropertyGraphStore interface does not currently expose full-text search directly — evaluate whether to add it as a custom method or via structured_query().

Implementation notes

• Full-text search in CoordiNode will be exposed as Cypher procedures: CALL db.index.fulltext.queryNodes(indexName, query)
• SDK should wrap this into a typed method that hides the Cypher procedure syntax
• hybrid_search() on the client can be upgraded to use full-text + vector as two legs of a single query

Gate

Blocked by CoordiNode DB tasks:

• R550: Full-text index build (tantivy integration in coordinode-storage)
• R555: Full-text Cypher procedures (db.index.fulltext.*)

Both are Phase 5.3. Do not start SDK implementation until a CoordiNode release with R550 + R555 is available.

Acceptance criteria

• client.full_text_search() returns results for text queries
• LangChain CoordinodeGraph.keyword_search() implemented
• hybrid_search() upgraded to combine full-text + vector scores
• Integration test with real CoordiNode instance
• Documented in README and API reference