Add checksum KAT to Github Workflow

[cothan]

This PR should be merged after #24 .

This PR add:

• gen_KAT.c: A way to deterministic generate KAT file.
• gen_NISTKAT.c: Generate NISTKAT file to compare with official/upstream ML-KEM KATs.
• checksum.sh: A bash script run executable and compare output of executable with an hash.
• Github workflow to test checksum of MLKEM512, 768, 1024 from 10,000 iterations.

• Does this PR change the input/output behaviour of a cryptographic algorithm (i.e., does it change known answer test values)? (If so, a version bump will be required from x.y.z to x.(y+1).0.)

[mkannwischer]

Please rebase on top of main and resolve any conflicts.

Do you think it's possible to switch to actual NISTKATs?
Unfortunately, the hashes in the kyber repo are not up to date, but you can generate them yourself from the code in https://github.com/pq-crystals/kyber/tree/standard.
You would have to re-implement it along the lines of https://github.com/PQClean/PQClean/blob/master/test/crypto_sign/nistkat.c - which should not be too hard. This will be much easier to match against the official testvectors that will be published by NIST togethr

MLKEM512:
4b88ac7643ff60209af1175e025f354272e88df827a0ce1c056e403629b88e04
MLKEM768:
21b4a1e1ea34a13c26a9da5eeb9325afb5ca11596ca6f3704c3f2637e3ea7524
MLKEM1024:
6471398b0a728ee1ef39e93bb89b526fbf59587a3662edadbcfc6c88a512cd71

[cothan]

Good idea. I will add NISTKAT test. I see it's no harm to add independent test as long as our checksum match NISTKAT.
The reason I didn't add it as first place was the AES part to generate random data, which can be done with FIPS202 (already included in XOF of MLKEM).
I want to separate the AES part out of the code, perhaps out of the repo so the actual code does not have any dependency with AES, do you have any idea?

We can always fall back to add a common/nistrng.c and a common/README.md say that nistrng.c is only used for NIST KAT testing.

[mkannwischer]

Thanks for making the changes. This looks good to me now.