Use default input sample rate for voice recording

[bricelb]

Summary

Use the default input device sample rate for voice recording instead of forcing 16 kHz.

Why

Some audio devices do not support opening an input stream at 16 kHz directly. On those systems, voice recording can fail even though the input device exposes a usable native sample rate such as 44.1 kHz or 48 kHz.

Changes

• Query the default input device's default_samplerate before opening the stream.
• Use that sample rate for InputStream, short-recording threshold calculation, and WAV metadata.
• Keep 16 kHz as a fallback when the backend does not expose a numeric device rate.
• Treat SSH as a notice rather than a hard warning, since audio can be local to the target machine.

Validation

• python3 -m py_compile tools/voice_mode.py
• git diff --check

[bricelb]

Superseded by #61463 — same fix rebased on current main, with tests added per the contributing guide.