Compare current runner artifacts with checked-out benchmark history

[tkilias]

Goal

Compare current downloaded benchmark artifacts with the historic benchmark results stored in the checked-out commit.

Scope

• Load the local checked-out benchmark-history directory as the baseline.
• Aggregate runner samples by median for each test set and benchmark.
• Flag regressions above a configurable threshold, defaulting to 10%.
• Report unmatched cases and runner samples separately.

Acceptance Criteria

• Comparison uses the history available in the current checkout, not the direct parent commit.
• Threshold, noise, regression, missing-baseline, and incomplete-candidate scenarios are tested.
• Unmatched data is reported without being classified as a regression.

Dependencies

• Depends on Define benchmark artifact, history, and comparison models #10, Store and load benchmark history collections #15, Normalize pytest-benchmark JSON for comparison #16, and Aggregate downloaded runner artifacts into a candidate collection #17.