Shrink size of CA issuance logs

Right now the CA fires off four very-similar log lines:
https://github.com/letsencrypt/boulder/blob/a0cb60ea6acd264b10fb9ddeeba517b32e503f27/ca/ca.go#L327
https://github.com/letsencrypt/boulder/blob/a0cb60ea6acd264b10fb9ddeeba517b32e503f27/ca/ca.go#L344
https://github.com/letsencrypt/boulder/blob/a0cb60ea6acd264b10fb9ddeeba517b32e503f27/ca/ca.go#L405
https://github.com/letsencrypt/boulder/blob/a0cb60ea6acd264b10fb9ddeeba517b32e503f27/ca/ca.go#L422

Each of these log lines contains a bunch of the same info: acctID, orderID, profile, and big blobs like the hex pubkey (inside the IssuanceRequest). But we could instead log all the redundant info once, and then tie the four log lines together with a single traceID-style unique identifier.

Decisions to make:

• is it worth deduplicating small info like the orderID?
• how do we want to deduplicate info that is in the embedded IssuanceRequest struct, which has other non-logging uses?

Start by reading the four cited log sites in ca/ca.go and the embedded IssuanceRequest struct. Compare which fields are repeated across the issuance logs, then define how they should be correlated and which details remain logged; the work is done when the four events retain the needed context with less redundant output.