Make draft private-key-format agnostic

[danvangeest]

With the disagreements going on about private key format for ML-KEM, can we make this draft format agnostic? I think the only places where this would apply is that we currently refer to ML-KEM.KeyGen() and ML-KEM.Decaps() from FIPS 203. These functions explicitly use the expanded key.