Miscellaneous bug fixes

[paulromano]

This PR has a few small bug fixes:

1. There's a fix for S(a,b) name suggestions instead of applying a guess #1945 so that you can pass table_name when caling ThermalScattering.from_njoy and it won't try to guess a name
2. A user noticed that when they tried to run a depletion calculation and specify cross sections using Materials.cross_sections (but not setting OPENMC_CROSS_SECTIONS) they ran into an exception. It turns out that the Operator._get_nuclides_with_data method always tried to find what nuclides have neutron data by just looking at the environment variable but not Materials.cross_sections. This has been fixed in this PR and I've tried to better encapsulate the logic for finding cross sections / depletion chains into single places (the _find_cross_sections and _find_chain_file functions).
3. There's a fix for model.run() can fail to return the statefile #1939

[pshriwise]

As I understand it, this doesn't quite fix the problem described in #1945.

The problem in the issue is that the user could not successfully reference the new thermal scattering data in material.add_s_alpha_beta because the name was not present in openmc.data.thermal._THERMAL_NAMES but the new name ("c_Graphite_20p") was close enough to an existing data name in _THERMAL_NAMES that the openmc.data.thermal.get_thermal_name function was returning a valid, but unintended, dataset by providing a close match. I've updated that issue with a little more info and proposed fix that we could include in a separate PR.

So, a matter of using already generated data rather than creating a dataset with custom names. Though the user did indicate in the forum thread that they also had to adjust some things script to ensure the right name was being applied to their data, so probably added some fixes similar to the ones you've applied here too.

[paulromano]

@pshriwise definitely did not realize the issue was with Material.add_s_alpha_beta. Now that I see that, I agree it doesn't make sense for us to second guess the user's choice here. I've added a commit that takes away the guessing in the add_s_alpha_beta method altogether.

[pshriwise]

This looks good to me @paulromano. Thanks!