Remove NumPy as a hard dependency

[dgasmith]

Description

This PR removes NumPy from opt_einsum as a runtime dependency. The change has little effect on the opt_einsum code base but significantly effects the testing infrastructure. Several testing strategies are currently employed to help understand the most effective one. Currently, these are:

• Move all NumPy functions to a testing.py module with pytest.skip
• Various pytest.skip functions wrapped in decorators
• Local NumPy imports.

Feedback welcome!

Todos

Notable points that this PR has either accomplished or will accomplish.

• Remove unused custom tensordot code
• Remove import numpy from possibly_convert_to_numpy
• Fix ssa_to_linear with a proper implementation

Questions

• Question1

Status

• Ready to go

[jcmgray]

LGTM so far! Do you understand the test failures, is that just an env thing?

[dgasmith]

I didn't get to every test yet. I'll hack a bit this weekend.

[dgasmith]

@janeyx99 Is this still of interest to Torch? I think we're relatively close on removing that last of the NumPy dependancies.

[janeyx99]

Yes! Removing the last dependencies would be awesome!

[dgasmith]

Apologies- I started a new job. This is in my mind and I hope to hack a little tomorrow night!

[dgasmith]

@janeyx99 I'm looking to push another release shortly, is this still a requested feature?

[janeyx99]

@dgasmith yes, torch still doesn't have a hard dependency on numpy (our tests do) so this is still desirable! Sorry it took a week to get back to you--this fell in my personal inbox.

[dgasmith]

I'll get in #232 first and then finalize this PR. It will be quite tidy to have no formal dependancies.