Update smirnoff_hack.py for OpenFF Toolkit 0.10.3

[mattwthompson]

We released a new version of the toolkit yesterday that broke smirnoff_hack by nature of adding a named argument to generate_conformers - fixing it should be a matter of string the new argument through the appropriate functions.

[mattwthompson]

CI seems to be failing for other reasons (pip and/or other deps changing). I might be able to have a look later.

[j-wags]

Ahh gosh, sorry, I keep forgetting about this file. Yes, the OFF Toolkit 0.10.3 release yesterday added a kwarg to the generate_conformers method signature which broke the caching logic here. I'm sorry for the inconvenience @leeping.

This code change looks correct to me. @leeping - ForceBalance's CI hasn't run since the 0.10.3 OFF toolkit release, so I can't show you the problem, but if we run CI on master right now it should fail because of this problem.

Apologies again, I think our ask is both for this PR and for a ForceBalance release. There may be other issues happening too with other deps and we could discuss how to handle those as well.

[mattwthompson]

I've updated this to be self-contained to the changes associated with the toolkit. I opened #249 in an effort to fix other failing tests. Hopefully that can be accomplished before this one.

[j-wags]

The approach from #231 (switching from explicit listing of args and kwargs to using *args and **kwargs) would probably be a more future-proof way of handling this.

In the long run it'd be great to fully move the caching out of forcebalance and into the toolkit. Maybe we could coordinate with @leeping on that so we can synchronize making that change and a subsequent release in the future? Then we could stabilize the FB interface and not cause this kind of problem in the future.

[mattwthompson]

Agreed, I've opened #251 with a smaller set of changes