Fix issues with thermal model handling

[jzuhone]

This PR fixes a number of issues related to ACISpy's handling for thermal models.

• The call to get_xija_model_spec in the function find_json to get the model specification file originally set check_version=True, but this function was often called many times in a short interval in some applications and github did not appreciate the repeated requests. So this option is now turned off.
• The find_json function also had some convoluted logic to look for the JSON file in a different chandra_models repository path which has now been fixed.
• The ThermalModelFromRun class "divided" a string by a string instead of using the divide operator to concatenate a Path object and a string. This has now been fixed.
• The ThermalModelRunner.from_backstop had an error in computing the initial temperature from the start of a model run. This has now been fixed, and required an update to one of the test answers.
• A new pytest fixture has been added to allow one to store new answers when running the tests if necessary.
• The interface with acis_thermal_check has been updated to match the changes made in Move all of the temperature model scripts into the main package acis_thermal_check#44.
• The planning limit for 1DEAMZT model was one degree lower than it should have been and this is fixed.

All regression tests pass with the answers fixed, and the new answers were hand-checked against the old. The new functionality to store answers has also been checked to ensure that the answers are identical if they are supposed to be the same.

[jzuhone]

@taldcroft does this need to go through the new control board process? it's not used for load review but it does get installed into skare.

[jzuhone]

@taldcroft just wanted to ask again if you thought this needed to go through FSDS (my supposition is not, but if you want to update it in skare you might have to?)

[taldcroft]

Sorry I missed the first comment @ mention. No, this does not need to go as a separate FSDS issue, but the assumption is that each PR that gets lumped into a Ska update (which goes to FSDS) has been reviewed internally, passes some sort of testing and does not break other stuff. We apply that to every Ska package update.

[jzuhone]

Ok--it's not necessary for this to go into the same ska update as acis_thermal_check if it's not convenient. It can wait until the next one.