fix: allow specifying language for audio transcription

[lahuman]

Summary

Add an optional channels.transcriptionLanguage setting so audio transcription can pass an explicit language hint to Groq and OpenAI Whisper requests.

Closes #2421.

What Changed

• add channels.transcriptionLanguage to the config schema
• propagate the transcription language through ChannelManager and BaseChannel
• pass language to both GroqTranscriptionProvider and OpenAITranscriptionProvider
• include the language multipart field in transcription requests when configured
• document the new config option in docs/configuration.md
• add tests covering config propagation and Groq request payloads

Why

Previously, transcription requests relied entirely on automatic language detection.

That works most of the time, but it can reduce accuracy when the detected language is wrong. This change allows users to provide an explicit language hint, which matches the behavior described in the issue and keeps the implementation small and configurable.

Configuration

Example:

{
  "channels": {
    "transcriptionProvider": "groq",
    "transcriptionLanguage": "en"
  }
}

[chengyongru]

Hi @lahuman

I have cherry-picked your modifications into the PR below. This makes the changes appear more complete—thank you for your contribution!

see: #3116

close for now