Come up with a better solution for how we handle claim_pattern

[mattmoor]

Right now we have to explicitly enumerate these in our schema because BigQuery doesn't support map[string]string. This leads to tedious changes like #89 when a users adds a new claim they want to match on (a losing battle).

Instead we should consider solutions such as serializing this field to a list of (name, value) pairs as we do for the actor claims.