Support Task-scoped QA commands

[ChinmayK0607]

Why

Task QA needs a scripting and CI surface in addition to the Platform UI. Users should be able to discover agents, run exact Task analyses, and consume stable outcomes without writing custom API calls.

What changed

• allow hud qa agents, run, and results to target Task subjects
• validate the selected agent contract before launching a run
• preserve exact selected attempts when polling and rendering results
• return quality and execution exit codes for terminal --no-wait results
• add regression coverage for Task scope, reuse, polling, malformed responses, and exit behavior

Result contract

• 0: passed
• 1: failed or unknown quality result
• 2: invalid request
• 3: execution, polling, or response-contract failure

Dependency

Stacked on #529, which adds the shared resource QA CLI.

Test plan

• uv run --extra dev pytest hud/cli/tests/test_qa.py -q (20 passed)
• Ruff
• Pyright

Made with Cursor