Surface review status when the runner selects a protocol

With #9, every protocol publishes structured reviews in PROTOCOLS.yaml — reviewer name, optional ORCID, date, the protocol_version reviewed, and a status from a controlled vocabulary (approved, verified-with-benchmark, changes-requested, deprecated). Nothing consumes it yet.

The protocol runner should read it and surface it at selection time, alongside the trust_tier and status signals it already uses (skills/protocol-runner/SKILL.md:29,36,41,84,143 in this repository).

Specifically:

• List reviews among the index fields the skill relies on.
• Show review status next to trust_tier and status in match listings and in the pre-execution summary.
• Warn when the version about to be executed is unreviewed — that is, when no reviews entry carries a protocol_version equal to the protocol's version. This is the case the format was designed to make visible: a protocol can carry several approvals of older releases while the current one has been vetted by nobody.
• Warn on changes-requested for the version being executed.
• Consider review status as a tiebreaker in ranking, after trust_tier and status.

Sequencing: this was waiting on the runner relocating out of bioconductor/ai-agent-skills and being renamed, so the change would land once rather than in two repos. That happened in #14 (merged 2026-09-11), so this is now unblocked and the paths above refer to its new home here.

Start with skills/protocol-runner/SKILL.md at lines 29, 36, 41, 84, and 143, then inspect the runner entry point and review fields in PROTOCOLS.yaml. Trace how trust_tier and status are shown and ranked. Done means review status appears in listings and the pre-execution summary, with version-specific warnings for unreviewed and changes-requested reviews.