ca/linter: Port safety checks from ceremony tool

[pgporada]

In #7005 several safety checks were added to the ceremony tool:

This change extracts the RawSubject to RawIssuer DER byte comparison into the //linter package proper so that it can serve both //ca and //cmd/ceremony.

Adds a helper function verifyTBSCertificateDeterminism to //ca similar to an existing check in //cmd/ceremony. This code is not shared because we want //cmd/ceremony to largely stand alone from boulder proper. The helper performs a byte comparison on the RawTBSCertificate DER bytes for a given linting certificate and leaf certificate. The goal is to verify that x509.CreateCertificate was deterministic and produced identical DER bytes after each signing operation.

Fixes #6965

[aarongable]

Logic LGTM, just some suggestions for cleanup

[aarongable]

a) don't directly check for nil here, instead check len(foo) == 0
b) no need for this since the anonymous function does the same check on line 673; pick one or the other

[pgporada]

What is the functional difference between checking for nil or len 0 on a []byte? The zero value of a []byte is going to be nil.

[aarongable]

You're right, the true zero value of []byte is nil. But generally the purpose of checks like this is not simply to check for the zero value, but rather to check for obviously invalid inputs, so that more helpful error messages can be displayed and so that we can avoid doing expensive checks if we already know they're going to fail. So a len(foo) == 0 check also catches the case where the byte slice exists but has an obviously-invalid (zero) length.

[pgporada]

I can get rid of this because tbs.Empty() on line 670 is handling the len(foo) == 0 case anyways.