Investigation bias rules don't get deleted based on counts

[matejminar]

A colleague brought attention to the fact that custom dynamic sampling rules for the investigation bias do not expire when the number of samples is exceeded. Had a short talk with @iambriccardo about this - it seems the current count is stored in Redis and accessed there by relay, and we don't want to transmit information back to sentry from relay. Checking Redis could be an option - but is it worth introducing feedback for something like this?