Add language input option / detected language in output / speaker identification

[Jeronymous]

• Add language selection (which makes sense with whisper STT) -- issue Add language selection #21
• Add speaker identification -- issue Add speaker identification #23

Note: this requires STT workers based on the new API supporting language (pull request linto-ai/linto-stt#48 -- issue linto-ai/linto-stt#53)

Documentation: search for "language" / "speaker identification" in the new README.md.

[Jeronymous]

In the API:

• the env variable LANGUAGE set the language for text normalization ("*" means no text normalization)
• (new) the configuration parameter language set the target language for Whisper STT ("*" means automatic language detection)
• (new) the results include a key "language" in all the text segments with the detected language (which is the language passed as an option, if it's not "*")

Corner case : what happens when the parameter language is not specified in the transcription config and the variable env LANGUAGE is set to something (ex: "fr-FR") ?
Currently, it passes nothing (language=None) to the STT worker, so the behaviour depends on the LANGUAGE variable env of the Whisper STT worker.
I think it's correct, but some users might think that the varianve env LANGUAGE of the transcription service is passed in that case. I'll try to clarify in the doc.