chore(ci): remove unused recordings

Added a script to cleanup recordings. While doing this, moved the CI
matrix generation to a separate script so there is a single source of
truth for the matrix.

Ran the cleanup script as:

```
PYTHONPATH=. python scripts/cleanup_recordings.py
```

We can eventually put this as part of the pre-commit workflow to ensure
that the recordings are always up to date and that no stale recordings
are left in the repo.

tests/integration/suites.py

@@ -180,3 +180,16 @@ SUITE_DEFINITIONS: dict[str, Suite] = {
         default_setup="ollama-vision",
     ),
 }
+
+# CI test matrix - single source of truth for continuous integration test configurations
+# This is used by:
+# - .github/workflows/integration-tests.yml (CI jobs)
+# - scripts/cleanup_recordings.py (unused recording cleanup)
+#
+# Each entry defines a (suite, setup) pair that runs in CI.
+# Note: Special test configurations (vllm weekly tests, manual inputs) are handled in the CI workflow.
+CI_MATRIX = [
+    {"suite": "base", "setup": "ollama"},
+    {"suite": "vision", "setup": "ollama-vision"},
+    {"suite": "responses", "setup": "gpt"},
+]