review-validator: make asynchronous timing wrappers shell-portable #89

Summary

Make review-validator's asynchronous timing guidance safe across supported executor shells.

Current State

review-validator/references/execution-timing.md recommends the declared shell's time -p. During a repository review on macOS with zsh, a fresh validator followed that guidance and produced zsh:1: command not found: -p; just ci never started. The runtime correctly retained a blocked launch, but recovery then required proof compilation and a quiescent graph.

Proposed Changes

Provide a runtime-owned, shell-independent execution/timing helper, or explicit tested wrappers for each supported shell. Document how to retain the actual subprocess exit status and monotonic elapsed time across asynchronous polling. Reject unsupported timing wrappers before launching a validation unit.

Benefits

Avoid failed validation launches and graph recovery overhead caused by the skill's own example while retaining reliable execution-side timing evidence.

Implementation Notes

• Cover zsh and bash natively on both macOS and Linux, plus cmd, Windows PowerShell, and PowerShell on native Windows.
• Keep the dispatched command, working directory, and environment unchanged.
• Test launch failure and nonzero command exits separately from timing-wrapper failures.
• Require passing native Linux and Windows CI for the implementation; local macOS results and a configured matrix are not substitutes for those results.
• Keep full Linux cloud/HPC provisioning and end-to-end environment verification in Provide and verify a Linux review environment for cloud and HPC #95. Linux timing CI alone does not prove the actual cloud computer or cluster is ready.
• Related: review-graph: default to compact runtime receipts and validation references #87, review-graph: schedule a coordinator lane without expected capacity failures #88.