For the moment, Go calls sendfile(2) to transfer at most 4MB at a time
while sendfile(2) actually allows a larger amount of data on one call.
To reduce system calls of sendfile(2) during data copying, we should
specify the number of bytes to copy as large as possible.

This optimization is especially advantageous for bulky file-to-file copies,
it would lead to a performance boost, the magnitude of this performance
increase may not be very exciting, but it can also cut down the CPU overhead
by decreasing the number of system calls.

This is also how we've done in sendfile_windows.go with TransmitFile.

goos: linux
goarch: amd64
pkg: os
cpu: DO-Premium-AMD
           │    old     │                new                │
           │   sec/op   │   sec/op    vs base               │
SendFile-8   1.135 ± 4%   1.052 ± 3%  -7.24% (p=0.000 n=10)

           │     old      │                 new                 │
           │     B/s      │     B/s       vs base               │
SendFile-8   902.5Mi ± 4%   973.0Mi ± 3%  +7.81% (p=0.000 n=10)

           │    old     │              new               │
           │    B/op    │    B/op     vs base            │
SendFile-8   272.0 ± 0%   272.0 ± 0%  ~ (p=1.000 n=10) ¹
¹ all samples are equal

           │    old     │              new               │
           │ allocs/op  │ allocs/op   vs base            │
SendFile-8   20.00 ± 0%   20.00 ± 0%  ~ (p=1.000 n=10) ¹
¹ all samples are equal

Change-Id: Ib4d4c6bc693e23db24697363b29226f0c9776bb0
Reviewed-on: https://go-review.googlesource.com/c/go/+/605235


LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Jorropo <jorropo.pgm@gmail.com>
Run-TryBot: Andy Pan <panjf2000@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@google.com>
Reviewed-by: Carlos Amedee <carlos@golang.org>