Variants provided:
* SEEC-Passthrough: Use raw entropy source output
* SEEC-PRF-TupleHash: SEE...

There's no reason why this needs to be done at initialization time
with the way we have this setup.

Since I recently wrote this for work, I might as well reuse it.

This is primarily to consolidate the imports since PQnoise pulls in
CIRCL, but on modern Intel s...

In theory this can trivially support all of the KEM suites provided by
CIRCL (including some hyb...

This was incomplete/partial due to lack of toolchain/standard library
support in the first place...

While I stand by my reasons for using the AGPL in the first place, it
doesn't do what I want it ...

I wrote a new one primarily to have a better Ed25519/ristretto255
implementation to use, and I a...

This should improve X25519 performance on non-AMD64 targets.

This should fix the non-AMD64 build.

This should avoid having to import the `dh` sub-package as much.

The noise-c PSK test vectors are skipped due to the protocol names being
old(?), and requiring a...

Being able to add new algorithms on the fly and have it tie into
`NewProtocol(s)` is probably us...

This is inspired by how Snow handles the PSK modifier. It's easier to
maintain the built-in pat...

People can just look at the `go.mod` if they care which libraries I use,
and if they don't trust...

No patterns use this, and that's even more nonsensical than a
pre-known `e` key.

Consistency etc etc etc.

* Prevent a rather unlikely nil deref in `HandshakeState.Reset()`
* `ErrDone` has a better mes...

This is non-standard, but I want to use this anyway, and it helps test
AEAD implementations that...

* Remove error checking for situations that are impossible with valid
patterns.

* Allow th...

While this passes the validity rules, it's stupid, and pointless.

This allows users to validate custom pattern definitions against the
specification's pattern val...

* Add extra notes regarding non-standard crypto functions.