Add encoders to top-level import list

[scotts]

Fixes #1017

Curiously, I tried writing a unit test for this, but it passed without making any changes. :/ So there's something with how we're importing things during test that is making torchcodec.encoders available even though it's not exposed in the top-level __init__.py.

However, when I loaded up an interactive python shell without this change, I got:

>>> import torchcodec
>>> import torch
>>> encoder = torchcodec.encoders.AudioEncoder(samples=torch.rand(1), sample_rate=16_000)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: module 'torchcodec' has no attribute 'encoders'. Did you mean: 'decoders'?

Same as the user. And then with this change:

>>> import torchcodec
>>> import torch
>>> encoder = torchcodec.encoders.AudioEncoder(samples=torch.rand(1), sample_rate=16_000)
>>> 

[Dan-Flores]

This is interesting, there is a subtle difference between how we import in tests: from torchcodec.encoders import AudioEncoder and how one can import in python: encoder = torchcodec.encoders.AudioEncoder(...)

[NicolasHug]

So there's something with how we're importing things during test that is making torchcodec.encoders available even though it's not exposed in the top-level init.py.

That's because torchcodec.encoders becomes available if we do something like from torchcodec.encoders import AudioEncoder (which we do everywhere in the tests):

In [1]: import torchcodec

In [3]: torchcodec.encoders  # not available
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[3], line 1
----> 1 torchcodec.encoders

AttributeError: module 'torchcodec' has no attribute 'encoders'

In [4]: from torchcodec.encoders import AudioEncoder

In [5]: torchcodec.encoders  # now available
Out[5]: <module 'torchcodec.encoders' from '/home/nicolashug/dev/torchcodec/src/torchcodec/encoders/__init__.py'>

If we really wanted to test the import, we'd have to spawn a separate "virgin" process whos imports aren't poluted. E.g. like in https://github.com/scikit-learn/scikit-learn/blob/eee6582df6f64678c81209037d3800b4f319fb88/sklearn/experimental/tests/test_enable_hist_gradient_boosting.py. Probably overkill.