Improve proxy performance

[stevapple]

In production, we found sshmux significantly impact SSH tunnel speed, which merely runs at 100Mbps while an NGINX forwarder can run at 1000Mbps (the full Internet bandwidth) under the same resource constraints.

It would be great if we can resolve the problem so that sshmux won’t be a bottleneck of speed.

[taoky]

I think that we could add an optional pprof first like taoky/ayano@b26dcc8, to understand where the bottleneck is.

[iBug]

Note that Nginx stream forwarding does nothing to the stream content and can leverage lots of speedup techniques like TCP offloading and userspace bypassing. On the other hand, sshmux has to decrypt-then-reencrypt the content.

Maybe try building with GOAMD64=v3 or v4, but I'm not sure how much it helps.