Add retries, repeated runs, and flaky-test reporting #7

Motivation

Intermittent failures are difficult to diagnose in CI. The runner should be able to retry failures intentionally while distinguishing a stable pass from a flaky pass.

Proposed work

• Add per-run and optionally per-test retry counts.
• Add a repeat mode for stress-running selected tests.
• Record every attempt and its duration.
• Report passed-after-retry tests as flaky rather than ordinary passes.
• Define exit-code behavior for flaky tests, including an optional fail-on-flaky mode.
• Include retry data in JSON and JUnit output.

Acceptance criteria

• CLI and programmatic APIs can configure retries.
• Attempt history appears in human and machine-readable reports.
• A deterministic fixture verifies pass, fail, and flaky outcomes.
• Retries interact correctly with bail and timeout settings.