Add warning level to DataValidator

[dc-almeida]

Closes #426
Adds a warning_level attribute to data validation YAML files and registers warnings in log, as specified in #426.
If present, failed validation does not raise errors for corresponding criteria.

[dc-almeida]

After discussion in weekly stand-up, will change the fail message start for the warning case, as it might be confusing to refer to a "failed" validation.

[danielhuppmann]

Looks good to me - a few suggestions inline, then good to be merged from my side.

[phackstock]

As just discussed bilaterally, I'd suggest a few changes detailed inline below.

[danielhuppmann]

Seems like I can't respond to the comment by @phackstock on the warning-level inline, so putting this here.

I think that the suggestion to use integers instead of strings is just confusing to non-expert users, without adding any practical benefit.

In yesterday's ScenarioCompass meeting, we discussed the need for different types of warnings in the sense of "warning for deviation from historical energy data" versus a "biodiversity concern in the future". Here, there is no clear ordering in terms of low/medium/high - instead, this could be clearly differentiated using colors.

Going one step further, it may be useful to define the allowed warning-level-names in the nomenclature.yaml project-config, not the package, so that we don't have to always make a new release when adding new levels.

[dc-almeida]

As discussed in our meeting, incorporating the following features now:

• warning levels, defaulting to error if not specified (ordering is implicit in the Enum)
• allowing multiple warning levels for same variable (to be refined in future)
• printing warning level in log output df's (this addition can lead to hiding other columns, such as 'variable', to preserve text length; see test_DataValidator_apply_fails; the variable is still printed in the criteria header, though, and row numbers ultimately identify error/warning sources)

Follow-up PRs will include:

• a way to refactor all warnings/errors into single df (ordered by warning level) for easy tracing and visualisation
• rewriting warnings for each variable such that only the highest level is printed (when multiple)
• include descriptions of constraints for warnings

[danielhuppmann]

Thanks, this looks good as a first step!

[phackstock]

Good to be merged from my side.