fix(feishu): include audio duration in uploads

[sgaofen]

Summary

• attach best-effort audio duration metadata when Feishu uploads voice messages
• probe duration from WAV headers first, then fall back to ffprobe or mutagen when available
• add regression tests for both upload-body duration plumbing and WAV duration probing

Root Cause

send_voice() reused the generic Feishu file-upload path, but the upload body never included audio duration metadata. Feishu's native audio message rendering depends on that field, so successful uploads could still display with missing or incorrect duration.

Closes #8300.

Validation

• python3 -m py_compile gateway/platforms/feishu.py tests/gateway/test_feishu.py
• uv run --extra dev pytest tests/gateway/test_feishu.py -q -k 'send_voice_uploads_opus_and_sends_audio_message or probe_audio_duration_uses_wav_header'
• uv run --extra dev pytest tests/gateway/test_feishu.py -q -k 'not test_build_event_handler_registers_reaction_and_card_processors'

Platform Tested

• macOS 15.x (Apple Silicon)

Contribution Guide Notes

• Reviewed CONTRIBUTING.md and checked for existing open PRs before submitting this scoped change.
• Ran the targeted verification commands listed above for this PR. I have not claimed a full repo-wide pytest tests/ -q pass unless explicitly noted.

[alt-glitch]

Overlaps with #16736 which also fixes #8300/#16524 (Feishu audio duration). Both PRs add mutagen/ffprobe probe chain. Recommend consolidating to avoid duplicate merge.

[teknium1]

Thanks for the focused Feishu regression coverage. The underlying gap remains on current main: plugins/platforms/feishu/adapter.py:4559-4563 builds audio upload bodies without duration, and tests/gateway/test_feishu.py:2527-2575 has no duration assertion.

Problems

• The diff targets gateway/platforms/feishu.py, but current main relocated that adapter to plugins/platforms/feishu/adapter.py in 552adbe0827c32df8ed9bb19e908c26eff43add7; the PR is currently DIRTY.
• The added probe performs WAV reads and potentially subprocess.run() synchronously inside the async upload path. Current adapter SDK work uses _run_blocking at plugins/platforms/feishu/adapter.py:1692-1695.
• The added ffprobe call has no timeout.

Suggested changes

• Port the change and tests to the plugin adapter/current test imports.
• Execute the probe through the adapter executor and give ffprobe a finite timeout while preserving the best-effort fallback.

Automated hermes-sweeper review.

[teknium1]

_probe_audio_duration_ms() can read files and launch ffprobe; this async upload path should await executor-backed work (for example the adapter's _run_blocking) rather than run the probe on the event loop.

[teknium1]

Please give this ffprobe invocation a finite timeout. Without one, a hung executable can indefinitely delay outbound voice delivery.

[teknium1]

Closing as stale target: this PR patches gateway/platforms/feishu.py, which was deleted when the platform adapters were rewritten as plugins — the live code now lives in plugins/platforms/*/adapter.py, so this diff can no longer apply.

Feishu upload duration is being salvaged via #40592 against the live plugin adapter — and credit where due: this PR was the earliest report/fix of the missing-duration problem.

Thanks @sgaofen — if anything in your change isn't covered by the salvage noted above, please resubmit against the current plugin adapter.