WIP: supports mutable IV in GcmParams, close #225

[zkonge]

so that some PKCS11 implementation (like AWS CloudHSM) could write random IV into it.

Fixes #225

[wiktor-k]

I'm not using this scheme personally but the underlying issue seems real and the trade-offs in this solution reasonable.

Wdyt @hug-dev @ionut-arm @keldonin et al?

[zkonge]

Although the trait constraint for Mechanism is enough for the alias rule:

Mechanism: !Clone+!Send

I open another branch 43c8863 that make all operations use the owned Mechanism.

It's not necessary, but I think it can give us (and the compiler) more restrictions on when we can get things wrong.

Any ideas about that? And is there anyone prefer passing the mutable reference instead of owned data?

[wiktor-k]

Any ideas about that? And is there anyone prefer passing the mutable reference instead of owned data?

Wait, I thought mutable ref would allow the PKCS11 implementation to set these fields that are to be read by the caller, which would be impossible when using owned data? (unless the owned object is being returned from the function too).

Btw, I think we can resume work on this given recent changes in main.

Sorry for the delays @zkonge 🙇