Add audio + video recording

[jlowin]

This PR starts to add tools for working with audio as easily as text (and video!). The marvin.audio module can record and play audio and is fully compatible with the marvin.ai.audio functions like transcribe and speak. marvin.video supports intermittent frame capture that is fully compatible with the marvin.beta.vision functions like caption and cast/extract/etc.

In order to record and play audio, the optional audio dependencies must be installed:
pip install "marvin[audio]"
Note that transcription / speaking is available without these dependencies, but Marvin can not be used directly for recording / playback.

For video:
pip install "marvin[video]"

Example of spoken conversation with an AI assistant:

import marvin
import marvin.audio

ai = marvin.beta.Application()

while True:
    # record the user and transcribe the audio
    user_audio = marvin.audio.record_phrase()
    user_text = marvin.transcribe(user_audio)
    print(f"User: {user_text}")
    
    # send text to AI assistant
    ai_response = ai.say(user_text)
    ai_text = ai_response[0].content[0].text.value
    print(f"AI: {ai_text}")
    
    # speak the AI response out loud
    ai_audio = marvin.speak(ai_text)
    ai_audio.play()

Live transcription:

import marvin.audio

recorder = marvin.audio.record_background()

# iterate over the audio clips
for clip in recorder.stream():
    print(marvin.transcribe(clip))

	if some_condition():
		recorder.stop()

Live captioning from webcam:

import marvin.video

recorder = marvin.video.record_background()

# iterate over frames
for frame in recorder.stream():
		print(marvin.caption(frame))

	if some_condition():
		recorder.stop()

[zzstoatzz]

super cool!

this closes #861

[darinkishore]

Hey, if anyone runs into an error (like me) while installing marvin[audio], run

pip install --upgrade setuptools wheel first.

Also, if you don't have portaudio installed, make sure it's installed, so python can build the PyAudio wheels! If on macos, brew install portaudio works.