WFE: Normalize names at the WFE

[beautifulentropy]

Modify the contract between WFE.NewOrder and RA.NewOrder to expect normalized names from the WFE. This adjustment is a part of a two-phase update, where RA will eventually stop normalizing names, avoiding redundancy.

Phase 1:

• [ ] Normalize at the WFE
• [ ] Fix WFE test RA.NewOrder mock to no longer normalize names

Phase 2:

• [ ] Stop Normalizing names at the RA

[aarongable]

(This came out of a comment on https://github.com/letsencrypt/boulder/pull/7554)

I'm not actually arguing in favor of this, I just said I thought we should think about it :)

It's not clear to me that it's a good idea to change the contract between the WFE and the RA -- the RA is where we have the vast majority of our safety logic, and that's a good norm to preserve. Also, there's basically zero cost to calling UniqueLowerNames twice, just in case.

My comment was more about:

• is it valuable to log the non-normalized names in the WFE log event?
• is it valuable to count the number of non-normalized names against the max names limit, rather than the normalized number of names?