Slightly improve spam filtering

[nonprofittechy]

We've just been getting one kind of spam recently, which appears to be mostly manual copy/paste on the feedback form. This updates the filter so it works on the current spam we're seeing.

I looked into bayesian spam filtering with antispam which did not catch this spam at all out of the box. rather than integrate that now, leaving it for a future exercise. probably getting good results will require training our own model on webform spam.

Opening up #54 to track this idea if we ever have capacity