Properly substitute trait generics

[Nadrieril]

Recent changes in charon made method generics explicitly represented by a binder. This removes the need to manually concatenate/split lists of generics. This PR updates eurydice to make use of this. It also seemed like associated types and parent trait clauses were used with incorrect generics so I fixed that. I checked that this does not change generated code for the tests.

[msprotz]

Yeah some weird errors in mlkem. I need to switch to something else today but can look at this early next week.

[Nadrieril]

I'm minimizing, it's clearly a subtitution error of some sort

[Nadrieril]

Minimized:

pub trait Hash<const K: usize> {
    fn PRFxN<const LEN: usize>(input: &[[u8; 33]; K]) -> [[u8; LEN]; K];
}

pub trait Variant {
    fn kdf<const K: usize, Hasher: Hash<K>>() -> [u8; 32];
}

pub fn decapsulate<Hasher: Hash<32>, Scheme: Variant>() {
    let _ = Scheme::kdf::<32, Hasher>();
}

error:

subtype mismatch:
  <N:0, CGS:1> size_t ->  uint8_t[33size_t]* -> uint8_t[$0][32size_t] (a.k.a. <N:0, CGS:1> size_t ->  uint8_t[33size_t]* -> uint8_t[$0][32size_t]) vs:
  <N:0, CGS:1> size_t ->  uint8_t[33size_t]* -> uint8_t[$0][$0] (a.k.a. <N:0, CGS:1> size_t ->  uint8_t[33size_t]* -> uint8_t[$0][$0])

the K in PRFxN is incorrectly resolved to LEN somewhere.