[9.0] fix: JobWrapper does not manage to send job parameters

[aldbr]

While trying to debug the issue we have with job parameters, I noticed we didn't have any way (or at least any obvious way) to change the log level of the job wrapper (INFO by default from what I can see).

BEGINRELEASENOTES
*WorkloadManagement
NEW: ability to adjust job wrapper log level
ENDRELEASENOTES

[fstagni]

Is "VERBOSE" already the default?

[aldbr]

Wait, it sounds like we can actually modify the LogLevel of the JobWrapper by either:

• changing the log level in the JDL of a given job
• or changing the log level of the JobAgent

I will give it a try

[aldbr]

After installing the branch in certification, I managed to investigate the issue properly and set up a reproducer.
Here is the problem:

• The CPUNormalizationFactor job parameter is submitted as a str in DIRAC.
• In diracx-core, we have a pydantic model for the JobParameters. The model expects to get CPUNormalizationFactor as an int. A validator is expected to convert CPUNormalizationFactor from str to int.
• Autorest is correctly generating the fields of the JobParameters but forget about generating the validators.

The JobReport.commit() was just silently failing. Even after adding some logs to it, the message is not crystal clear but it helped me debug the problem 🙌

[aldbr]

I made a fix that I need to test properly in the certification machine again just to make sure it was the only problem.

[aldbr]

In DIRACGrid/diracx#653 we said that CPUNormalizationFactor should be treated as a float.
I still process it as an int here, because I would also need to modify diracx and OS before adding it to FLOAT_PARAMETERS.
Once merged I will open an issue to keep track of it of course.

[aldbr]

Tested and fixed in certification: you can have a look to the job parameters of job 10179