test_relative_setup_charge_change failure

[mikemhenry]

CI has a failing test: https://github.com/choderalab/perses/actions/runs/3101942658/jobs/5023771528#step:9:322

         old_system_charge_sum = sum([old_nbf.getParticleParameters(i)[0].value_in_unit_system(unit.md_unit_system) for i in range(old_nbf.getNumParticles())])
        new_system_charge_sum = sum([old_nbf.getParticleParameters(i)[0].value_in_unit_system(unit.md_unit_system) for i in range(new_nbf.getNumParticles())])
        charge_diff = int(old_system_charge_sum - new_system_charge_sum)
>       assert np.isclose(charge_diff, 0), f"charge diff is {charge_diff} but should be zero."
E       AssertionError: charge diff is -1 but should be zero.
E       assert False
E        +  where False = <function isclose at 0x7fb46cc7f820>(-1, 0)
E        +    where <function isclose at 0x7fb46cc7f820> = <module 'numpy' from '/usr/share/miniconda/envs/test/lib/python3.9/site-packages/numpy/__init__.py'>.isclose

[ijpulidos]

Apparently we need to pin the OpenmmForceFields version as well, until we can support the latest OpenFF toolkit version.

[ijpulidos]

There are two problems in the latest CI. One with openff.units imports and another which is the one directly linked here. They might be related.

[mikemhenry]

Solved with #1124