[shaohaozhi286]

Fix the Linux Desktop build failure introduced by speech recognition: the WebKitGTK audio/video permission queries are named webkit_user_media_permission_is_for_audio_device and webkit_user_media_permission_is_for_video_device. Preserve the existing local-origin and audio-only checks.

Add a small Linux CI job that compiles the production native Desktop source against real WebKitGTK headers before the full C++ build finishes. This would catch the invalid symbol names without waiting for all native dependencies and test sources to build.

The immutable v1.0.10 tag points to the failing source and is preserved. Prepare v1.0.11 with the full release notes for users upgrading from v1.0.9; no v1.0.10 Release assets were published. Update the existing speech OpenSpec record and keep hardware microphone validation explicitly unverified.

Validation:

• Workflow YAML parses and git diff --check passes.
• Exact failure evidence: package run 39151062016, linux-x64 job 117744826641; PR Release ACECode v1.0.10 #145 native build failed at the same stage.
• Linux native API compilation passed in run 39152273460 / job 117748175419 (45 seconds). Static and Web checks also passed; the C++ full regression job continues. The follow-up commit only records this evidence in tasks.md and does not change compiled code.
• Existing Web tests/build, 102 refactor tests and seed revision/fingerprint checks passed in Release ACECode v1.0.10 #145; Web source and seed contents are unchanged in this PR.

API reference: https://webkitgtk.org/reference/webkit2gtk/2.36.3/WebKitUserMediaPermissionRequest.html

GitHub publication only; aupdate mirroring is excluded by the user's explicit instruction.