SPARQL endpoint queries empty graph — wire to project ontology

[JohnRDOrazio]

Problem

The SPARQL endpoint (POST /api/v1/search/sparql) always executes queries against an empty RDFLib Graph, returning empty results for every query. The SPARQLQuery schema already defines ontology_id and default_graph fields, but they are completely unused.

Root Cause

• The route handler calls service.execute_sparql(query) without passing a graph
• The service silently falls back to Graph() (an empty graph) when no graph is provided
• No ontology is ever loaded for the SPARQL endpoint

Expected Behavior

When ontology_id is provided in the request, the endpoint should load the corresponding project's ontology graph and execute the SPARQL query against it, returning actual results.

Proposed Solution

See docs/plans/sparql-ontology-loading.md for the full implementation plan. Key steps:

1. Load the ontology graph in the route handler using the existing _load_graph pattern from quality.py
2. Add branch resolution (default to project's default branch)
3. Add authentication/access control
4. Extract the shared _load_graph helper to avoid duplication
5. Make ontology_id required (or validate in the route)

Files to Modify

• ontokit/api/routes/search.py — add db/user/branch deps, load graph
• ontokit/schemas/search.py — make ontology_id required or validate in route
• ontokit/services/search.py — optionally add service-level guard
• Extract shared _load_graph helper from duplicated pattern