[Audio::transcribe] Add support for timestamp_granularities

[jplhomer]

What:

• [ ] Bug Fix
• [x] New Feature

Description:

This PR adds support for the new timestamp_granularities[] property in Audio Transcriptions.

• Adds the ability to describe the granularity level of the transcription
• Processes the response to include words
• Adjusts the HTTPTransporter Payload class to support sending array-based data

Currently, attempting to pass in a granularity level array results in an error:

$response = OpenAI::audio()->transcribe([
  'model' => 'whisper-1',
  'file' => $file
  'response_format' => 'verbose_json',
  'timestamp_granularities' => ['word'],
]);

Exception:

First argument to "Http\Message\MultipartStream\MultipartStreamBuilder::createStream()" must be a string, resource or StreamInterface.

My solution is to parse one level of array data. I'm not sure if this is the best solution, or if we ought to support recursive array parsing.

[solomonjames]

This looks great overall, and worked for me once I made that small tweak to your code. Thank you for pushing this up, I hope we can get it merged into the main project, as I have it forked just to use this feature.

[gehrisandro]

Resolve here https://github.com/openai-php/client/pull/374

Thanks @jplhomer for your work!