RA: Count new registrations with contacts

[beautifulentropy]

Adding a temporary metric to estimate the rate of new accounts that include a contact.

Part of #7966

[aarongable]

LGTM. Might want to also increment the metric (perhaps with a different label) in the updateAccount codepath, right before calling ra.UpdateRegistrationContact.

[beautifulentropy]

LGTM. Might want to also increment the metric (perhaps with a different label) in the updateAccount codepath, right before calling ra.UpdateRegistrationContact.

I’ve actually slightly improved the implementation after noticing that the number of contacts permitted is variable and I chose to increment inside of UpdateRegistrationContact and explicitly after ra.validateContacts has returned without error. We only want to include contacts that we would actually be willing to dispatch.

[aarongable]

What's the purpose of having both "new" and "update" labels, when they are always set to have opposite values?

[aarongable]

LGTM with two minor notes:

• Feels a little odd to do the metric increment after writing to the SA in NewRegistration, but do it before writing to the SA in UpdateRegistrationContact. Makes readers question if there's a reason for the inconsistency. Sorry I didn't notice this on first pass.
• Doing this after ra.validateContacts is great for now, but I don't necessarily want to set the precedent that we'll be validating contacts in boulder before exporting them. I think that other systems are much better equipped to handle that than we are (beyond basic basic sanity checks). So far I've been imagining that in the future, contacts will be exported from the WFE, and never make it to the RA at all.

[jsha]

Since this is now counting "number of contacts received" rather than "number of registrations with contacts," the metric name and PR summary should be updated to match.

[beautifulentropy]

Since this is now counting "number of contacts received" rather than "number of registrations with contacts," the metric name and PR summary should be updated to match.

Hah, good call. I hadn't noticed the incongruity post-change.