Add an automated test suite for the JSONL replay and parsing logic

[aasis21]

Enhancement

There are currently no automated tests in the repo. The JSONL edit-log replay engine (set_path, append_path, replay_jsonl) and parse_chat_model are intricate and easy to regress when handling new VS Code chat formats.

Suggested improvement

Add a unit test suite (stdlib unittest or pytest) with small fixture files covering:

• kind 0/1/2 JSONL replay edge cases (nested set, list splice, out-of-range index).
• parse_chat_model extraction of turns, files, and refs.
• file_uri_to_path decoding.
• FTS query escaping (ties into FTS query crashes on special characters in search text #1).

This locks in behavior and protects against format drift.

Affected

• New tests/ directory; engram.py parsing functions