Add host fields and protos for ssh identities

[fspmarshall]

Updates sshca.Identity to support host identity fields and adds SSHIdentity to the decision family of protos.

The original sshca.Identity object was added in #50787 and was initially specialized to user identities only. At the time I thought I'd be able to mostly avoid caring about host identities. It turns out that there is enough ssh certificate handling logic that is intended to work with host and user certs, that it is best to be able to handle both using the same identity type.

I'm not completely happy with having host and user identity fields living alongside one another, and I did experiment a bit with the idea of having separate HostIdentity and UserIdentity types, with a common interface so that logic that is generic over both can continue to operate on identities as a single object. This ended up being a bit too much scope creep for the purposes of these changes (especially since the more common tlsca.Identity is already monolithic).

Part of ongoing work related to the PDP proposal (https://github.com/gravitational/teleport.e/pull/5736).

[mvbrock]

LGTM, just a couple questions out of curiosity

[mvbrock]

Just curious what the value of explicitly rejecting the request is when the valid before/after is set.

[mvbrock]

And would this make more or less sense to put in the req.Check() method? More just asking for my own edification.

[fspmarshall]

The reason is that the values are intended to be derived from TTL by this function rather than supplied by the caller, so a nonzero value in one of those fields may indicate human error. Basically, this check is just intended to help catch mistakes.

In terms of the location of the check, no real reason... I was just working on this part of the code when I first decided that it would be good to have a check like this. Moved it to req.Check().