[codex] fix baseline ci failures

[CarlosArmeroMoneo]

What changed

This PR fixes the baseline CI failures that were already present on main.

The patch is intentionally small and avoids touching the live OVH real-execution path:

• removes the existing Ruff violations in benchmark scripts, persistent_executor, and a sidecar test
• fixes the MyPy issues in validation_confidence, campaign_metrics, embedded_session_client, persistent_client, and persistent_executor
• updates the test_real_slice prewarm test so the tiny_network path no longer depends on an installed qiskit package in core CI

Root cause

main was failing for three separate reasons:

• stale Ruff violations in files unrelated to the recent wording cleanup
• a handful of deterministic MyPy typing issues
• tests/test_real_slice.py::test_real_executor_emits_prewarm_provenance[tiny_network] imported qiskit.QuantumCircuit inside the prewarm path, but the core CI job does not install the quantum extra

Impact

• restores the baseline static checks
• keeps the OVH VM campaign behavior unchanged by leaving src/aqs/execution_real.py runtime logic alone
• makes the prewarm provenance test self-contained in CI

Validation

Local validation in the clean worktree:

• python -m ruff check src tests scripts
• python -m mypy src/aqs
• python -m pytest -q tests/test_real_slice.py -k "prewarm_provenance"
• python -m pytest -q tests/test_validation_confidence.py
• python -m pytest -q tests/test_tiny_mnk_sidecar.py
• python -m pytest -q tests/test_embedded_session_client.py

Note: tests/test_persistent_executor.py is AF_UNIX-based and does not run on this Windows host. That coverage is expected to come from GitHub's Linux CI runner, which matches the failing workflow environment and the OVH deployment model.