KEM interface tweaks

[rozbb]

Made ergonomic changes to the KEM trait so that people can implement their own KEM. All of the newly exposed functions and fields are #[doc(hidden)] because this is really for power users.

Fixes #23 #25 #26

/cc @BoOTheFurious what do you think?

[codecov-commenter]

Codecov Report

Merging #27 (83b5e9c) into master (5ff4b00) will decrease coverage by 0.24%.
The diff coverage is 90.90%.

@@            Coverage Diff             @@
##           master      #27      +/-   ##
==========================================
- Coverage   92.72%   92.47%   -0.25%     
==========================================
  Files          15       15              
  Lines         742      744       +2     
==========================================
  Hits          688      688              
- Misses         54       56       +2     

Impacted Files	Coverage Δ
src/serde_impls.rs	96.96% <ø> (ø)
src/setup.rs	100.00% <ø> (ø)
src/single_shot.rs	100.00% <ø> (ø)
src/kem.rs	90.00% <50.00%> (-10.00%)	⬇️
src/kem/dhkem.rs	91.54% <93.47%> (+0.12%)	⬆️
src/kat_tests.rs	95.65% <100.00%> (+1.44%)	⬆️
src/kdf.rs	88.88% <100.00%> (ø)
src/op_mode.rs	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5ff4b00...83b5e9c. Read the comment docs.

[BoOTheFurious]

That's great ! Nevertheless I still do not understanding why sk_eph is of type PrivateKey (#24 ). With this name it should be a SharedSecret or perhaps another dedicated type (or simply aVec<u8>). There is no direct link between the ephemeral secret object and the private key (except in some specific cases like DHKEM)

Thanks for the very quick work !

[rozbb]

Oh, good point. Hmm, I'm leaning towards making it a hidden associated type, but I'll have to think about it a little. Will get back to this soon.

[rozbb]

I figured, rather than create a new EphemeralKey type, it'd be much cleaner if I just made implementors impl encap() alone. In reality, the only reason encap_with_eph exists is for testing purposes. So I moved it away and into kat_tests::TestableKem.

Let me know if this interface seems reasonable.

[BoOTheFurious]

It seems great to me !

[rozbb]

Cool, please let me know if you're able to get an initial version of your PQ KEM working. If so, and everything works fine, I'll cut a new version

[BoOTheFurious]

👍 it works for me, Big thanks

[rozbb]

Just cut a new version! Feel free to use 0.9 as your dep.