Adding initial tutorial

[vincentqb]

Initial tutorial with an overview of dataset loading and transformers. Created using the following notebook. See demo.

Things that may or may not make it in this PR:

• Add as a follow-up a part on migrating to pytorch from kaldi, and kaldi compliance support.
• Update as soon as the transformers have been standardized, see PR. For instance, we expect LC2CL to no longer be needed.
• Confirm location to put tutorial
• Link to a code/notebook/lab, as done here.
• Find a better solution to correct the path to images than adding a command in Makefile.

In a later iteration, we should demo a neural network implementation, maybe inspired by a prior version (which no longer appears in the left hand side menu but is still searchable).

[vincentqb]

Since there is an ongoing discussion about dataset abstraction, I replaced VCTK for one of our test file, converted from mp3 to wav.

[vincentqb]

Discussion with @jamarshon and @cpuhrsch.

• Let's expand the introduction, and talk about how the library in general is useful and leverages GPUs.
• In the kaldi section, let's showcase fbank and spectrogram. Some of the others could be made private.
• torchaudio is the official spelling 😄

[vincentqb]

I talked with @nairbv and he suggested putting the tutorials here. See PR to import into tutorials.