feat(dashboard): set run experiment and tags - #1490

Summary

• Add Experiment and Tags controls to the Dashboard Run Eval modal.
• Send experiment and tags in the run-launch API request using the existing snake_case wire contract.
• Validate experiment names and tags server-side before spawning the eval process.
• Pass --experiment to the CLI and write initial tags to the new run workspace's tags.json sidecar.
• Keep resume/rerun flows from changing experiment or setting initial tags on existing runs.

Notes

This branch was rebased after PR #1489, so new Dashboard-created runs now follow .agentv/results/<experiment>/<timestamp>/ through buildDefaultRunDir().

No dogfood screenshots or browser evidence are committed to this public repo. Any future UAT evidence should be stored on an agentv-private evidence branch.

Worker validation

The worker reported these checks on the original branch before the artifact-layout rebase:

• bun test apps/dashboard/src/components/run-eval-threshold.test.ts
• bun --filter @agentv/sdk build
• bun test apps/cli/test/commands/results/serve.test.ts -t "POST /api/eval"
• bunx biome check ...changed TS files...
• bun --filter agentv typecheck
• bun --filter @agentv/dashboard build
• Browser UAT with agent-browser on desktop/mobile modal flows

GitHub Actions is the merge gate for the rebased branch.