Implement PKCS12 Export in terms of Pkcs12Builder for non-Windows

[vcsjones]

Currently, the UnixExportProvider is implemented by putting together ASN.1 data in a bit of an ad-hoc fashion using the serializers.

In order to support #80314 and #111560, we will soon want to offer the capability to export in PKCS12 with different configurations for PBES, like PBES2-AES256-SHA256.

Rather than try to tack on PBES2, PBKDF2, etc in to the ad-hoc encoder, this pulls in Pkcs12Builder in to System.Security.Cryptography. Because it is currently in System.Security.Cryptography.Pkcs, we can't use it directly because it is a package, and even if we ship the package, we have a circular reference problem.

To address that, this puts Pkcs12Builder and its dependencies in Common. This is, essentially:

1. Move the files. We use the existing BUILDING_PKCS define to change the visibility. If we are building the public API, its visibility is public, otherwise it is internal.
2. PkcsHelpers needed to be split.
3. Some things in SignedCms needed to be moved to PkcsHelpers because we don't want to pull in SignedCms.

Contributes to #80314.

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-security, @bartonjs, @vcsjones
See info in area-owners.md if you want to be subscribed.

[vcsjones]

Why do we need to suppress these if we are in System.Security.Cryptography?

Because S.S.C.Pkcs targets netstandard, the analyzer that tells you to use the throw helpers does not kick in. However when this is pulled in to S.S.Cryptography, the analyzer lights up because it is unaware the source file is being pulled in to a non-net project.

[vcsjones]

Enumerable.Empty is in System.Linq. S.S.Cryptography does not have a reference to Linq, and it didn't seem worth it for this one case. An empty collection expression is probably "right" these days anyway.

[vcsjones]

Do we need any unit tests for this, Kevin?

I am open to suggestions. Export, directly and indirectly, is used considerably in existing tests to the point where "keep it working" seemed sufficient. I did make a small mistake when bringing this up initially. My computer was probably close to catching on fire from all of the test failures and debug asserts I hit.