🐛 Make RL sampling and Qiskit actions user-seedable

[flowerthrower]

🤖 AI text below 🤖

Description

Makes RL training-circuit sampling, seeded inference, and Qiskit's randomized
passes reproducible only when the user supplies a seed.

On a fresh environment, reset() uses Gymnasium's entropy-seeded generator for
circuit sampling and passes no seed to Qiskit. Native pass settings therefore
remain unchanged, including SABRE's CPU-dependent trial defaults.
Predictor.train_model(seed=...) and PredictorEnv.reset(seed=...) opt into
reproducible mode: they control circuit selection and supply derived native
seeds to QiskitSabreMapping, SabreSwap, VF2Layout, and VF2PostLayout.
Later seedless resets continue the initialized Gymnasium RNG stream, matching
Gymnasium and Stable-Baselines3.

Seeded SABRE execution uses one native layout and swap trial so results do not
depend on CPU-count-based defaults. Candidate QASM paths are sorted before
indexed selection. compile_as_predicted retains stochastic policy and native
pass behavior by default; supplying seed=... selects deterministic masked
argmax inference and reproducibly seeds randomized Qiskit actions.

This does not add the repeated figure-of-merit-aware wrapper tracked in #667,
change BQSKit's established fixed pass seed, or alter unrelated ML estimators.

It depends on #796 and is followed by #798. No
new package dependencies are introduced.

Part of #664

Validation after removing AI routing: all 101 non-training compilation tests pass on the combined stack, and full lint passes. The existing action invariants cover the remaining actions. Fresh CI for this rebased head is pending.

Checklist

• The pull request only contains commits that are focused and relevant to this change.
• I have added appropriate tests that cover the new/changed functionality.
• I have updated the documentation to reflect these changes.
• I have added entries to the changelog for any noteworthy additions, changes, fixes, or removals.
• I have added migration instructions to the upgrade guide (if needed).
• The changes follow the project's style guidelines and introduce no new warnings.
• The changes are fully tested and pass the CI checks.
• I have reviewed my own code changes.

If PR contains AI-assisted content:

• Any agent that created, edited, or submitted GitHub content was explicitly authorized for that scope, as required by our AI Usage Guidelines.
• Every agent-authored or agent-edited public text body begins with the visible disclosure 🤖 *AI text below* 🤖 (titles are exempt).
• I have disclosed AI assistance in the PR description.
• I confirm that I have personally reviewed and understood all AI-generated content, and accept full responsibility for it.

[codecov]

Codecov Report

❌ Patch coverage is 85.71429% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/mqt/predictor/rl/helper.py	0.0%	2 Missing ⚠️
src/mqt/predictor/rl/actions/qiskit_actions.py	90.9%	1 Missing ⚠️
src/mqt/predictor/rl/predictor.py	85.7%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!