Transition Minutes from API to library

[iKevinY]

• [x] Methods to add audio files corresponding to speakers (lazily)
• [x] Method to train final layers on the above speaker information

This is the usage that I'm currently envisioning:

from minutes import Minutes

m = Minutes()

m.add_speaker('speaker-1', '/audio/speaker-1/a.mp3')
m.add_speaker('speaker-1', '/audio/speaker-1/b.mp3')

m.add_speaker('speaker-2', '/audio/speaker-2/a.mp3')

# Perform transfer learning, training last layer(s) with the above data
m.train()

# Do prediction
m.predict(X)

[chadlagore]

Might I suggest,

m.predict('/audio/conversation.mp3')

[chadlagore]

Handled in #107