feat(settings): selectable audio input devices in settings

[milaforge]

This change adds audio input device selection to the settings UI so WhisperTux can record from non-default sources, including PipeWire/ALSA inputs such as Android microphones.

It keeps the implementation narrow:

• exposes available input devices in the existing settings dialog
• validates the selected device before saving it
• applies the device to the running audio capture instance without requiring a restart
• normalizes legacy config values like "default" and stringified device IDs
• adds focused regression tests for device-reference normalization and switching

Why:

• fixes cases where PortAudio’s default input is not the source the user wants
• avoids persisting broken audio-device settings
• improves readability and alignment in the related settings rows without broader UI churn

Validation:

• python3 -m unittest discover -s tests
• python3 -m compileall main.py src tests
• git diff --check

Notes:

• linter tooling was not available in this environment
• If two input sources share the exact name and host_api, restore is still ambiguous.