Implement benchmark comparison and history storage CLI commands

[tkilias]

Goal

Implement the artifact aggregation, comparison, and history-storage CLI behavior.

Scope

• compare consumes downloaded artifact bundles plus a local history root.
• compare is read-only and exits nonzero only for regressions or invalid input.
• store validates a complete candidate collection and writes it as the current revision history collection.
• Neither command runs pytest or benchmarks.

Acceptance Criteria

• CLI tests cover aggregate directories, current-checkout comparisons, output format, and failure exit codes.
• Store rejects incomplete candidate collections.
• Compare does not modify stored history.

Dependencies

• Depends on Add benchmark CLI entry point #13, Store and load benchmark history collections #15, Aggregate downloaded runner artifacts into a candidate collection #17, and Compare current runner artifacts with checked-out benchmark history #18.