normalize FAERS counts using data in OHDSI on actual prescription rates

[rkboyce]

we actually should add some data to drug_hoi_evidence that normalizes the FAERS data using actual prescription data. Counts and PRRs within FAERS can be meaningless because when the relative rate of prescribing is not known. Given two drugs with the same number of FAERS reports for a given HOI, the one that is less often prescribed is likely to be more of a concern. We can easily add that data now and should.