bump version to 0.6.5

Performance release: faster init / reprocess / increment with no graph,
schema, or CLI changes. Bulk COPY FROM graph writes (#341-#342) and the
lifespan-cached LayeredIgnore (#340) take init from ~395s toward ~140s on
the profiled medium corpus; the graph-write phase drops ~316s -> ~0.4s.

No re-index required -- the bulk path is byte-equivalent to 0.6.4 (verified
node-for-node and edge-for-edge, all properties + GraphMeta counters).
Ontology version unchanged (17).

Co-authored-by: Claude <noreply@anthropic.com>

Author: HumanBean17

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "java-codebase-rag"
-version = "0.6.4"
+version = "0.6.5"
 description = "MCP server for semantic + structural search over Java codebases"
 readme = "README.md"
 requires-python = ">=3.11"