Store Hashes instead of Tokens

[jbspeakr]

Current Behaviour

Hashes of tokens are calculated "on-the-fly" during the GET /revocations. This means not just that the tokens are stored plaintext in the DB, but also that there could be an increased resource usage if the GET /revocations result-set is large –– I already saw huge (>10s) latency spikes.

Expected Behaviour

Hashes are calculated and persisted during POST.