[geaflow/ai-memory] Harden keyword index serialization golden tests

Priority: P0
Difficulty: Starter

Context: Keyword retrieval exists through Lucene/keyword vectors, but the main branch behavior has been described as rebuilding during query and vulnerable to noisy common tokens.

Scope:

- Add golden tests for keyword tokenization, serialization, reload, Unicode, stopwords, and common-token pollution.
- Ensure incremental add/delete expected behavior is documented.

Constraints:

- Do not implement a new keyword engine.
- Coordinate with any existing resident keyword index PR.

Acceptance Criteria:

- Golden tests fail if tokenization or serialization changes unexpectedly.
- Common tokens do not dominate exact graph anchoring fixtures.

Suggested paths:

- `geaflow-ai/src/test/java/org/apache/geaflow/ai/index`
- `geaflow-ai/src/test/resources/index/keyword`

Start in geaflow-ai/src/test/java/org/apache/geaflow/ai/index and inspect the existing keyword-index tests and related fixtures under geaflow-ai/src/test/resources/index/keyword. Run the relevant tests before changing them, then cover tokenization, serialization and reload, Unicode, stopwords, common-token pollution, and incremental add/delete expectations. Done means the golden tests catch unexpected changes and common tokens do not dominate the exact graph anchoring fixtures.

Written by the indexing model from the issue text.