`trait ValidatedRsaPrivateComponents` over `struct RsaPrivateKey`

[pinkforest]

The current RsaPrivateKey is just a validated construct essentially.

Are there any reasons on

a) why it can't be just expressed as a trait benefiting no-alloc impl ? e.g. why do we need hold copies ?

b) why that can't be handed over to end-user e.g. hazmat unchecked or validating PKCS#8 wrapper ?

Given no-alloc will require memory efficient representation for it

• [WIP] Fixed-size no_alloc-friendly keys #636

type PKCS8Rsa impl VerifiedRsaPrivateKey
  type [..]: Unsigned // Can be optimised across exp & modulus over largest variant of say just one generic `U`
  ..

Pros:

• hazmat End-users can provide unchecked / checked construct as-is without specialised API surface
• Generalize impls around that in no-alloc and alloc
• Could provide conditional for Multi Prime (MP-RSA)
• Could provide conditional for precomputed
• Optimised memory pressure given usage-scenario/s
• Direct PKCS#8 (or other containers) borrowed layout into validated type
• Associated types can be optimized for it's intended use in algorithms

Cons:

• Complexity ?
• Harder ?
• .. ?