test: strengthen durable correctness coverage for responses state

[YueZh127]

Closes #667

Problem

Responses durable 主链已经有基本覆盖，但在 #667 关注的几类边界上，测试还不够扎实：

• LlmSession 对 forwarded tool call 的权威状态推进还缺少更完整的 durable contract 覆盖
• ResponsesAgentToolState 对 terminal / expired state 的稳定读取与恢复语义覆盖不足
• host /v1/responses 对 expired / cancelled / terminal state 的错误表达还缺少更细的边界测试
• continue / tool result / terminal rejection 的一些路径，之前更多覆盖 happy path，缺少“不能继续推进”的锁定

Solution

这次 PR 只补测试，不改生产语义、不改公开接口，围绕 #667 增加 durable correctness coverage：

• 在 LlmSessionGAgentTests 中补 forwarded tool call durable lifecycle 覆盖
• 补 terminal state 的吸收态与重复 resolve 幂等语义
• 在 ResponsesAgentToolStateGAgentTests 中补 terminal / expired state 的稳定查询与恢复语义
• 在 ResponsesAgentToolStateCurrentStateProjectorTests 中补 current-state projector 对 terminal state 的对外呈现保护
• 在 MainnetResponsesEndpointsTests 中补 host 级 durable rejection coverage，确保 cancelled / expired / terminal state 不会被错误继续推进，也不会错误调用 provider

Scope

本 PR 仅包含测试改动，未修改生产代码。

涉及文件：

• test/Aevatar.GAgentService.Tests/Core/LlmSessionGAgentTests.cs
• test/Aevatar.GAgentService.Tests/Core/ResponsesAgentToolStateGAgentTests.cs
• test/Aevatar.GAgentService.Tests/Projection/ResponsesAgentToolStateCurrentStateProjectorTests.cs
• test/Aevatar.Hosting.Tests/MainnetResponsesEndpointsTests.cs

Key Coverage Added

• LlmSession forwarded tool call 从 Pending -> Received -> Resolved 的 durable 状态闭环
• session expired 后 forwarded tool call 保持 expired，不被后续 signal 误复活
• cancelled forwarded tool call 保持 terminal，不再接受 result
• duplicate resolve 不推进 version，也不覆盖第一次 ResolvedAt
• ResponsesAgentToolState 在 expired / cancelled / resolved 等 terminal state 下可稳定查询
• host 在 previous_response_id expired / cancelled 时返回结构化错误，而不是继续 provider path
• host 在 cancelled / expired tool call 下返回 tool_call_not_available，且不写入 tool result、不调用 provider

Validation

已执行：

dotnet test test/Aevatar.GAgentService.Tests/Aevatar.GAgentService.Tests.csproj --nologo --filter "LlmSession|ResponsesAgentToolState|ResponsesCompletion"
dotnet test test/Aevatar.Hosting.Tests/Aevatar.Hosting.Tests.csproj --nologo --filter "MainnetResponsesEndpointsTests"
bash tools/ci/test_stability_guards.sh
bash tools/ci/query_projection_priming_guard.sh
bash tools/ci/projection_state_version_guard.sh
bash tools/ci/projection_state_mirror_current_state_guard.sh

结果：

• Aevatar.GAgentService.Tests 目标过滤通过
• MainnetResponsesEndpointsTests 通过
• test stability guard 通过
• projection / query 边界相关 guards 通过

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.51%. Comparing base (5d7162f) to head (0df3352).
⚠️ Report is 115 commits behind head on dev.

@@            Coverage Diff             @@
##              dev     #742      +/-   ##
==========================================
+ Coverage   82.32%   82.51%   +0.19%     
==========================================
  Files         932      941       +9     
  Lines       59485    60101     +616     
  Branches     7805     7872      +67     
==========================================
+ Hits        48970    49594     +624     
+ Misses       7155     7114      -41     
- Partials     3360     3393      +33     

Flag	Coverage Δ
ci	82.51% <ø> (+0.19%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.
see 53 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.