STT Audio Formats & Dialect setting

[chughts]

Copy of relevant points from feature request for Node SDK.

STT now supports audio/mp3 and audio/mpeg for all methods that accept an audio format.

Custom language models now have a new field: dialect. For en-US and ja-JP models, this will always match the name of the language (en-US or ja-JP). For es-ES models, this will be es-ES by default but can also be es-LA or es-US if specified by the user.

The new dialect field is optionally specified with the CustomModel object that is passed as part of the POST /v1/customizations method and is always returned as part of the Customization object by the GET /v1/customizations and GET /v1/customizations/{customization_id} methods.

[chughts]

mp3 and mpeg format checks added in 0.5.19