Gate legacy kitless rendering on PRs

[nvsekkin]

Description

Makes legacy kitless rendering failures fail PR checks while preserving full test execution; merge after OVPhysX wheelhouse parity is restored.

[greptile-apps]

Greptile Summary

The PR changes CI gating for kitless rendering:

• Makes the legacy rendering variant blocking on pull requests.
• Keeps the ovstage variant non-blocking on pull requests.
• Keeps both variants non-blocking for manually dispatched workflows.

Confidence Score: 5/5

The PR appears safe to merge, with the workflow condition matching the declared rendering-gate behavior.

The matrix contains the expected legacy and ovstage variants, and the new condition correctly makes only legacy failures blocking on pull requests while leaving manual runs non-blocking.

Important Files Changed

Filename	Overview
.github/workflows/build.yaml	Updates continue-on-error to gate legacy rendering failures on pull requests while preserving the intended non-blocking behavior for OVStage and manual runs.

_{Reviews (1): Last reviewed commit: "Gate legacy kitless rendering on PRs" | Re-trigger Greptile}

[isaaclab-review-bot]

Isaac Lab Review Bot

The workflow change makes legacy kitless rendering failures block pull requests while preserving non-blocking manual runs and existing push behavior.

• Design and architecture: The gating matches the matrix configuration: pull requests exclude the ovstage variant, so legacy is the only kitless variant that becomes blocking on PRs. The separate rendering-correctness job is unaffected.
• API: No Python, CLI, packaging, or public API surface changes. As this only modifies CI workflow behavior and touches no package, no package changelog fragment is needed.
• Implementation: The event and matrix paths are consistent: PR legacy runs are blocking, PR ovstage runs remain excluded, both workflow-dispatch variants tolerate failure, and push runs remain blocking. The ovstage PR condition is currently unreachable but remains consistent if the matrix exclusion changes later.

No blocking issues. No inline issue met the actionable-evidence threshold; the assessment above records the review feedback.

Automated review; human maintainers own approval decisions.