test(a2a): assert TaskResult content for rejected and auth-required states

[bokelley]

Closes #263 (item 3 of 5 — test coverage gaps for A2A 1.0 states)

Summary

• Adds test_rejected_task_result_content: asserts TaskResult shape for TASK_STATE_REJECTED — status=SUBMITTED, data=None, metadata["status"]="rejected", text message from TextPart.
• Adds test_rejected_task_adcp_error_datapart_not_extracted: documents the known gap — a REJECTED task's adcp_error DataPart is silently dropped because _process_task_response only calls _extract_result_from_task for COMPLETED tasks. The test asserts current behavior (data=None) and comments the future fix path.
• Adds test_auth_required_task_result_content: asserts TASK_STATE_AUTH_REQUIRED returns status=SUBMITTED, data=None, metadata["status"]="auth-required", TextPart message, plus task_id and context_id in metadata.

Prior tests for these two states (test_task_id_retained_on_auth_required_state, test_task_id_cleared_on_rejected_state) only asserted adapter.active_task_id; these tests close the TaskResult content gap called out in the issue.

Items NOT in this PR (flagged for @bokelley — see triage comment on #263)

• Item 1 (wrapper policy — docs say MUST reject, Python unwraps silently): cross-repo normative conflict + potential behavior change → human decision needed
• Item 2 (webhook envelope wrapping — bare vs StreamResponse-wrapped): requires A2A 1.0 §4.3.3 spec read before any code changes
• Item 4 (integration test for 1.0 client → 0.3 server): no 0.3-only server fixture exists; significant infrastructure needed
• Item 5 (CHANGELOG 4.1.0 subclasser breakage): CHANGELOG.md is bot-managed (release-please); add to MIGRATION_v3_to_v4.md but @bokelley should confirm scope

What was tested

• pytest tests/test_protocols.py — 76 passed (3 new + 73 existing)
• ruff check src/ — clean

Known nits (not fixed — surface for reviewer)

• result.success is True for TASK_STATE_REJECTED in the else-branch of _process_task_response — semantically wrong but pre-dates this PR
• The gap-documenting test could carry a pytest.mark.xfail marker once a fix is planned; left as a comment for now to keep this PR minimal

Pre-PR review

• code-reviewer: approved — no blockers; nit on docstring wording and gap-documenting test completeness (both addressed)
• dx-expert: approved — flagged typo in prompt diff (not in actual code; confirmed clean); nit on context_id assertion in auth test (added)

Session: https://claude.ai/code/session_01MxALGppjocjuZzDVp48tps

---

Generated by Claude Code

[github-actions]

IPR Policy Agreement Required

Thank you for your contribution! Before we can accept your pull request, you must agree to our Intellectual Property Rights Policy.

By making a Contribution, you agree that:

• You grant the Foundation a perpetual, irrevocable, worldwide, non-exclusive, royalty-free copyright license to your Contribution
• You grant a patent license under any Necessary Claims
• You represent that you own or have sufficient rights to grant these licenses

To agree, please comment below with the exact phrase:

I have read the IPR Policy

You can read the full IPR Policy here.

---

I have read the IPR Policy

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}