General Exception handling for /compute/output/{task_id}

[coltonbh]

Update 2025-02-08

I need to more robustly handle exceptions in this endpoint.

CASES:

1. When a qcop exception does not have e.program_output. E.g., when requesting a TeraChem calculation but TeraChem isn't available in BigChem. fr.get() raises a qcop.exceptions.ProgramNotFoundError but e.program_output is None. So None gets returned to the client as the "result."
2. Non qcop exceptions such as a parser error. I guess this has more to do with qcop properly wrapping up qcparse errors.

I need to either capture ALL exceptions in qcop and append a .program_output object to them, or handle this some other way. I don't love depending on lower packages to not have 500 errors in this endpoint, but otherwise it's a bit hard (impossibel?) to construct a program_output object for the end user because I don't get have the .input_data unless it's passed up by the underlying programs.

Exceptions raised in qcop, e.g., sending a "transition_state" calculation to "psi4" which doesn't support it will result in 500 errors because I don't generically handle exceptions on this endpoint.

The issue is that in qcop I choose to raise certain exceptions upfront before the program is ever called and do not create exc.program_failure objects for them. So these exceptions get raised in ChemCloud and I do not have the associated .input_data to create a proper ProgramFailure objects.

I think the best fix will be to handle ALL exceptions in qcop the same way and append a .program_failure object to them.