Improve error_code, make HDD1 errors be warnings

[elad335]

• Allow setting error as warning level so HDD1 errors would be logged as warnings.

Implements #12667

[Nekotekina]

I would like to not add new thread_local variable. I think it should be solved similarly to not_an_error: possibly a constructor with additional argument which transforms error_code report into a warning or something.

[elad335]

Hmm, I'll remove the method to manually set it. But I want to prevent multiple reports in the same syscall and the ability to silence errors that should be ignored. (so wrapping methods in CellError instead of error_code for example can be avoided)

[Nekotekina]

I'd suggest passing a reference to logs::message to make_report.
This would allow to choose both channel and severity:
return {sys_fs.warning, CELL_ENOENT, path};

[Nekotekina]

Where do you encounter double reports? It's usually easily avoidable, but moving report to GPR binding:

1. Creates a chance of missing report in certain cases (sub-calls) where it could be desirable.
2. Slower function exit for errorless case

[elad335]

Removed for now.