Prepare PQC KEM KAT Tests for Classic McEliece

[FAlbertDev]

This PR relates to the generalized PQC KEM KAT tests introduced in #3807. The following changes have been made.

• Refactor the PK_PQC_KEM_KAT_Test class to use polymorphism instead of a templated Delegator.
• Prepare the class for Classic McEliece, i.e., allow the child class to decide which KAT values are hashed in the KAT files. Also, allow the child class to override a callback to modify how the Fixed_Output_RNG is checked after keygen/encaps. In Classic McEliece, we do not know how much randomness is required; therefore, we need to soften the check in this case.
• Apply the changes to the Tests of Kyber and FrodoKEM

[coveralls]

coverage: 92.009% (-0.003%) from 92.012%
when pulling 0e30809 on Rohde-Schwarz:refactor/pqc-key-kat-test
into b8afed8 on randombit:master.

[reneme]

Nice. Certainly, this is more ergonomic than the delegate template. I'm still trying to remember why I went the template route initially: I believe, I didn't see a way to abstract from the different algorithm "mode" objects and didn't even think of using the high-level API in pk_algs.h. Thanks for spotting that!

A few things, but generally LGTM.

[FAlbertDev]

Thanks for your review, @reneme! I applied your suggestions.