(feat) Add Playwright testability hooks to the agent playground

[mmabrouk]

The agent playground UI has almost no data-testid attributes. During the v0.105.0 release-night QA (2026-07-14), manual UI QA had to pixel-hunt every control, and automation isn't practical as a result.

Spec

A complete spec already exists: docs/design/agent-workflows/projects/qa/playwright-testability.md on branch qa-prerelease-agent-runtime. It defines:

• Tier 1: ~10 always-on data-testids. Top 5 blocking: approval dock, tool call row, chat message, session status, commit approval.
• Tier 2: a debug-mode hidden node (behind NEXT_PUBLIC_AGENTA_TEST_HOOKS=1) exposing warm/frames/harness/sandbox/turn timing.

Ask

Implement the spec so the manual golden-path QA can become a Playwright CI suite.

Todo

• Add the Tier 1 always-on data-testid attributes (approval dock, tool call row, chat message, session status, commit approval, plus the remaining ~5 from the spec).
• Add the Tier 2 debug-mode hidden node behind NEXT_PUBLIC_AGENTA_TEST_HOOKS=1.
• Convert the manual golden-path QA into a Playwright CI suite using the new hooks.

[linear-code]

AGE-3957

[arushiiii18]

I'd like to work on the Tier 1 testability hooks if this issue is available.