Generating charges in the tests

[jthorton]

I noticed that we generate partial charges in some protocol tests, such as test_dry_run_default_vacuum. In this case, we repeatedly generate the same charges for each sampler type, which is an expensive operation when the charge generation itself is not what is being tested. I think it would be best to have a single explicit test which checks that a protocol is correctly generating charges, and then for all other dry runs to use pre-calculated charges to keep the tests quick, maybe this is something we should come up with a best practice for and add it to the dev guide we plan on making?

[IAlibay]

Re-opening PR #1131 - this dealt with it, but we ended up closing it for unknown reasons.