Initial prototype for scheduler parsing

[espenfl]

Here is an initial approach which enable basic scheduler parsing. The Slurm scheduler is used as an example. For Slurm, the State entry in the sacct contains for instance OUT_OF_MEMORY when the process breach the set memory limit. As such it makes sense to parse the sacct information and in case needed, map that to generic error codes.

There is also a method to parse the stdout and stderr of the parser, which in this commit does nothing. Currently the error code with the highest exit_status number is set (in case there are differences between the sacct and file approach.

Currently, exit codes are set on JobCalc. We might, in the future consider to move the same framework onto the scheduler framework to allow for specific handling of scheduler error codes without relying on specific definitions in a given JobCalc.

A test is also included, which sets up a node and tests the parsing for the sacct. We should, when functionality added, also add tests the file parsing.

[sphuber]

Thanks a lot @espenfl great first start, I have a few comments. Also the logic on which exit code to give priority should probably be discussed, in person would be most efficient I think

[ramirezfranciscof]

Hey @espenfl , could you briefly explain to me the relationship between this PR and #3261? They would seem to be related (the other one seems more broad but also seems to be dealing with parsing the scheduler). Which is the current status of each?

[sphuber]

Superseded by PR #3906 and #3931