Minor features and improvements

[juanmc2005]

This PR addresses issue #57.

Changelog

• Remove IrregularFileAudioReader
• Remove file audio readers
• Improve profiling implementation, there's now a profile() function that wraps an observable and a sequence of operators into an observable that tracks the processing time of each item
• Remove possibility not to emit the corresponding waveform alongside the each prediction
• Move feature pre-calculation from the pipeline to a new PrecalculatedFeaturesAudioSource that emits audio chunks and their corresponding segmentation and embeddings
• In diart.stream, catch the event of the user closing the plot window and end execution properly
• Make diart.blocks a package instead of a single file and group classes into categories for better organization
• Clean some code