Transcription with Timestamps - wrong parameter type in python example

Hi, 

it seems that there is a type mismatch in the python example on

docs/capabilities/audio_and_transcription.md

regarding Transcription with Timestamps:

Instead of 
`timestamp_granularities="segment"`

it should read

`timestamp_granularities=["segment"]`

as a list is expected.

