Store and load benchmark history collections

[tkilias]

Goal

Persist and load complete benchmark collections in Git-trackable history.

Scope

• Store packaged artifacts grouped by revision, test set, and runner.
• Preserve every raw runner sample and its metadata.

Acceptance Criteria

• Multiple test sets and multiple runner samples round-trip without lost associations.
• Incomplete and duplicate collections fail clearly.
• Stored layout conforms to the versioned artifact contract.

Dependencies

• Depends on Define benchmark artifact, history, and comparison models #10 and Package raw pytest-benchmark output as a portable runner artifact #14.