Prove / lint that filters are in a 'restrictive' order.

[gregglind]

Claim: showing that the survey deployment rules are in a topo sorted order is the best we can do here.

Complications: regexes, probabilities, etc. We cant show if geo=us or channel=release is more restrictive without probabilities. we can say they are indeterminate