Add scenario grouping and targeted EventLab replay

[elkampu]

Problem

The v1 manifest is intentionally a flat corpus inventory. That is enough to verify isolated adapter behavior, but replaying the whole manifest can make unrelated cases look like one trace and cannot express an ordered lifecycle or reconnect scenario.

Examples that need grouping include a Frigate new → update → end lifecycle and an ONVIF delivery followed by a reconnect duplicate.

Desired outcome

Add a small, bounded way to group existing case IDs into named ordered scenarios, plus targeted EventLab replay. Please discuss the manifest shape before implementing it so v1 fixtures remain useful.

Acceptance criteria

• Existing v1 case verification remains valid, or any version change has an explicit migration path.
• A scenario references registered case IDs in a deterministic order and cannot bypass corpus/path/size limits.
• replay --scenario <id> emits only that scenario; replay --case <id> emits one case.
• Waited replay preserves relative timing without sleeping before the first selected case; --no-wait remains deterministic.
• Unknown selectors, duplicate references, empty scenarios, and excessive scenario size fail before stdout is written.
• Schema, CLI help, tests, and fixture documentation describe the behavior.
• The compatibility matrix continues to report fixture evidence without implying that unrelated cases form a physical-device trace.

This is maintainer-created roadmap work and does not count as external adoption or an external behavior contribution.