crypto: document that the GCM AEADs are safe for concurrent use if the underlying cipher.Block is

[cespare]

This is a follow-on to #25882.

That issue is about documenting which cipher.Block implementations are safe for concurrent use.

Similarly, the cipher.AEAD created by NewGCM and friends ought to be documented as safe for concurrent use if the underlying cipher.Block is. In particular, AES-GCM should be safe.

/cc @FiloSottile

[FiloSottile]

Sounds good, send a CL with a test if you'd like! Thank you!

[awnumar]

chacha20poly1305 should also either be documented as concurrent safe or not.

From a glance at the code it appears as though it is safe to use concurrently as the only state that's accessed in the Open/Seal functions is a read of the key, though there are also assembly implementations I don't understand.

[gopherbot]

Change https://go.dev/cl/799066 mentions this issue: crypto/cipher: document that GCM AEADs are safe for concurrent use