Document default ASR engine

[hey-august]

The cloud platform uses the Deepgram speech-to-text engine by default.

We should describe this and link to their documentation, particularly the language codes

We also need to document the way language name and code are used by ai.languages.

• The name field is used in the system prompt so the AI knows which voice to use for a given language
• The code field is used for ASR and should match a code from Deepgram's language docs linked above

Sections to be updated

• Supported Voices and Languages
• ai.languages