[feature request] output a word level time stamp result into csv or json file

[eisneim]

First of all thank @ggerganov for creating such a useful project! super helpful!

it would be nice if the program can output a json file that contains detailed info of each word's timestamp

why is this useful? because now we can edit the text to edit the video, the APP "descript" has this feature, people editing podcast or talking video would be very fast!

the option --output-words output script for generating karaoke video is ok, but it's hard to extract the timestamp