Mechanism to filter out known errors

[mthuurne]

When using third party frameworks and components, getting errors fixed might take a while. To make APE's reports more relevant, it would be useful to have a mechanism by which known problems can be suppressed, so regressions will stand out.

Static code checkers often allow adding error suppression comments to source code, but that isn't going to work here, since we're checking the output and not the source. Suppression markings in the output doesn't really make sense, since if it's easy to customize the output, it's probably easy to fix the error as well.

The simplest possible implementation would be a list of error message texts. Perhaps regex support would be useful. Maybe some suppressions should only be applied to certain plugins or certain parts of the URL namespace. Even if we don't implement those features from the start, we can pick a file format that will support them.

[mthuurne]

The HTML validator supports message filtering. However, if we let the validator filter, we only have the option to not see a particular message at all; if APE does the filtering we could also allow downgrading of the message, where it doesn't mark a page as failed but is still included in the report. That approach would be more appropriate for errors that are real but aren't easy to fix.

[mthuurne]

For messages generated by APE or APE plugins, it would be nice to have a mechanism to disable them using a technical name instead of the message string, similar to how pylint has a technical name for each rule.