treewide: allow multiple validators

[msanft]

This changes the attestation (as of now, only SEV-SNP) to be passed
multiple validators. The aTLS code already handles multiple validators,
but the code previously passed only one. This way, attestation will now
work by being handed a list of validators, and returning success as soon
as one can successfully validate a report. Furthermore, the
atls.NoValidator is now obsolete, and semantically represented by
passing an empty list of validators.

[burgerdev]

lgtm

[Freax13]

I'm not quite done with the review yet, but here are some preliminary comments.

[Freax13]

Could you add some tests please?

[msanft]

Could you add some tests please?

I'm not exactly sure how we would go about testing the validators. There were no unit tests before, and implementing some would definitely require some considerations. I don't think this is something that should be done in this PR.

For end-to-end testing, we'd probably also want a scenario where we'd have multiple reference values. But with the current Azure offerings, I don't think it's possible to request such a platform.

What did you have in mind exactly @Freax13 ?

[Freax13]

For end-to-end testing, we'd probably also want a scenario where we'd have multiple reference values. But with the current Azure offerings, I don't think it's possible to request such a platform.

I was thinking that we could have a test where we have multiple reference values with only one of them being the correct one and then make sure that attestation still works. If we really wanted multiple different platforms, we could also generate such reports once and hard-code them for the tests.

[msanft]

For end-to-end testing, we'd probably also want a scenario where we'd have multiple reference values. But with the current Azure offerings, I don't think it's possible to request such a platform.

I was thinking that we could have a test where we have multiple reference values with only one of them being the correct one and then make sure that attestation still works. If we really wanted multiple different platforms, we could also generate such reports once and hard-code them for the tests.

Okay, I guess we could have some kind of testdata and put this into a unit test, or have an e2e test