some possible improvements to encrypted/internal/e2c/e2c.go

[cathugger]

ive noticed
https://github.com/Arceliar/ironwood/blob/main/encrypted/internal/e2c/e2c.go

func Ed25519PrivateKeyToCurve25519(pk ed25519.PrivateKey) []byte {
	h := sha512.New()
	h.Write(pk.Seed())
	out := h.Sum(nil)
	return out[:curve25519.ScalarSize]
}

not performing any of the usual scalar masking as done in https://github.com/teserakt-io/golang-ed25519/blob/master/extra25519/extra25519.go#L16-L26 for example.
it should probably be done as per https://neilmadden.blog/2020/05/28/whats-the-curve25519-clamping-all-about/
it's probably ~FINE because curve25519.ScalarMult used by nacl/box does it on its own but honestly it's not exactly clear whether all users of this value will do that, so I suggest doing it explicitly for brevity.

also Ed25519PublicKeyToCurve25519 should probably just use https://pkg.go.dev/github.com/teserakt-io/golang-ed25519/extra25519#PublicKeyToCurve25519 or https://pkg.go.dev/filippo.io/edwards25519#Point.BytesMontgomery instead of messing with bigints because bigints are kinda eh. likely slower and not constant-time (yes i know that derivering stuff from pubkey doesn't exactly require constant time).
also upstream this part was copied from changed as per FiloSottile/age@e63c22e

or you can just ignore this stuff because current code is kinda okay