ChaCha20Poly1305 is slow on ARM64

[zx2c4]

While amd64 has some amount of accelerated code for ChaCha20 and Poly1305 (chacha20poly1305_amd64.s), ARM64 has none. This has been a shortcoming for some time, but is now resulting in more and more bug reports to me as users complain about WireGuard on mobile being battery inefficient [android, ios], since all the crypto uses Go.

Somebody posted to gerrit an implementation of chacha20 and another person posted to gerrit an implementation of poly1305, but there's been no timeline set for merging of these, and I haven't seen any public plans posted about pursuing alternative implementations either.

This issue is to track the status of ChaCha20Poly1305 acceleration on ARM64, so that perhaps it can be added to some milestone or something.

[zx2c4]

Ahh, this appears to be a duplicate of #22809 - closing.