docs: Reorder track queue and initialize final stabilization tracks

- Initialize asyncio_decoupling_refactor_20260306 track

- Initialize mock_provider_hardening_20260305 track

- Initialize simulation_fidelity_enhancement_20260305 track

- Update TASKS.md and tracks.md to reflect new strict execution queue

- Archive completed tracks and remove deprecated test performance track

conductor/tracks/mock_provider_hardening_20260305/spec.md

@@ -0,0 +1,14 @@
+# Specification: Mock Provider Hardening
+
+## Background
+The current `mock_gemini_cli.py` provider only tests the "happy path". It always returns successfully parsed JSON-L responses, which masks potential error-handling bugs in `ai_client.py` and `AppController`. To properly verify the system's robustness, the mock must be capable of failing realistically.
+
+## Objective
+Extend `mock_gemini_cli.py` to support negative testing paths, controlled via an environment variable `MOCK_MODE`.
+
+## Requirements
+1. **MOCK_MODE parsing:** The mock script must read `os.environ.get("MOCK_MODE", "success")`.
+2. **malformed_json:** If mode is `malformed_json`, the mock should print a truncated or syntactically invalid JSON string to `stdout` and exit.
+3. **error_result:** If mode is `error_result`, the mock should print a valid JSON string but with `"status": "error"` and an error message payload.
+4. **timeout:** If mode is `timeout`, the mock should `time.sleep(120)` to force the parent process to handle a subprocess timeout.
+5. **Integration Tests:** New tests must be written to explicitly trigger these modes using `ApiHookClient` and verify that the GUI displays an error state rather than crashing.