I'm using both gcc and MSBuild parsers for different projects.
I think different teams would be using different settings here, as they may have different views on the severity of stylistic warnings.
I think there is no need to make this parser specific. There is already a setting "Categories to ignore", this feature would add "High/Medium/Low Categories" and/or replace that by a mapping table where users can override default classifications with "High/Medium/Low/Ignore".
It might also be useful to some people to have a similar mechanism for matches against file names or warning messages and/or combinations of these (e.g. demote certain warnings in third-party components without hiding them), but this would probably increase the complexity quite a bit.