Using uplink splice to reduce CPU time

[JkLondon]

Hello there!
On one of the heavily loaded servers (amd64 btw) I did a pprof check that showed that around 11% of CPU time was spent on the uplink buf.Copy.

I investigated this and found that we can cut the time by using splice for the uplink directly but @Fangliding rejected my PR (in the PR you can see the proposed implementation), saying that the expected performance change would be small (also you mentioned that a Linux/arm64 build could be used on Android but I think we can deal with it). To support the discussion I've attached the profiling graph below.

So, to sum up, 2 questions:

1. Are you considering fixing the TODOs (reader and writer)?
2. What should be changed to make uplink splice acceptable?