Enhance server-based speech recognition and configuration options

[74th]

This pull request refactors the audio format and language handling for speech recognition throughout the codebase to centralize configuration and simplify interfaces. It also adds support for using a Whisper server as a speech recognition backend and introduces a startup script for the Whisper server. The most important changes are summarized below:

Centralization of Audio Format and Language Configuration

• Replaced individual sample_rate_hz, channels, sample_width, and language_code parameters with a centralized LISTEN_AUDIO_FORMAT and LISTEN_LANGUAGE_CODE in stackchan_server/listen.py, stackchan_server/speech_recognition/google_cloud.py, and stackchan_server/speech_recognition/whisper_cpp.py. All speech recognition classes and methods now use these shared settings, reducing code duplication and potential for misconfiguration. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17]

Whisper Server Integration

• Added WhisperServerSpeechToText as a new speech recognition backend, integrated into the app selection logic in example_apps/echo.py. The app can now use a remote Whisper server if the appropriate environment variables are set. [1] [2] [3]

Startup Script for Whisper Server

• Added misc/whisper-server/run-whisper-server.sh to simplify running the Whisper server with parameters sourced from environment variables.

Improvements to WhisperCppSpeechToText

• Made model_path optional and fallback to the STACKCHAN_WHISPER_MODEL environment variable, improving configuration flexibility and error handling. [1] [2]

These changes collectively make the audio handling more robust and consistent, and enable easier deployment and configuration of speech recognition services.