[FIX] Do not charge baseline trials against the optimization n_trials budget (+ test)

[VincentG1234]

Summary

Baseline runs should not consume the configured optimization trial budget (n_trials). The synchronous path in _run_baseline_trials still incremented completed_trials, which caused the study to stop early (e.g. an incomplete grid). This now matches _collect_completed_trials, which already counts only optimization trials.

Changes

• auto_tune_vllm/core/study_controller.py

  • Removed completed_trials += 1 after each baseline finishes (success, failure, timeout, exception).
  • Docstring clarifies that n_trials applies to optimization trials only.

• tests/core/test_study_controller.py (generic module for future StudyController tests)

  • Fake backend that completes a trial on the first poll.
  • Test: two entries in concurrency_levels → completed_trials stays 0 after _run_baseline_trials.

Expected behavior after merge

• You can set n_trials (and grid cardinality in grid mode) without manually adding +1 / +k for baselines.
• Multiple concurrency_levels still mean multiple baseline Optuna trials, but they do not eat the optimization budget.

How to test

• pytest tests/core/test_study_controller.py