Add optimized ReadAll for http responses

[rdner]

What does this PR do?

1. When the body content length is known it's much faster to pre-allocated the entire buffer once.
2. When the length is unknown using bytes.NewBuffer + io.Copy is much faster.

Also, added benchmarks to prove the difference.

Why is it important?

When we start using this function in main products it will significantly improve the performance.

Checklist

• My code follows the style guidelines of this project
• I have commented my code, particularly in hard-to-understand areas
• I have added tests that prove my fix is effective or that my feature works

Benchmark

goos: darwin
goarch: arm64
pkg: github.com/elastic/elastic-agent-libs/transport/httpcommon
BenchmarkReadAll/size:_100/unknown_length/io.ReadAll-10                       	    1000	      1232 ns/op	     576 B/op	       3 allocs/op
BenchmarkReadAll/size:_100/unknown_length/bytes.Buffer+io.Copy-10         	    1000	      1219 ns/op	     224 B/op	       4 allocs/op

BenchmarkReadAll/size:_100/known_length/io.ReadAll-10                     	    1000	      1275 ns/op	     576 B/op	       3 allocs/op
BenchmarkReadAll/size:_100/known_length/bytes.Buffer+io.Copy-10           	    1000	      1206 ns/op	     176 B/op	       3 allocs/op

BenchmarkReadAll/size:_102400/unknown_length/io.ReadAll-10                	    1000	     60858 ns/op	  514405 B/op	      24 allocs/op
BenchmarkReadAll/size:_102400/unknown_length/bytes.Buffer+io.Copy-10      	    1000	      9833 ns/op	  106640 B/op	       8 allocs/op

BenchmarkReadAll/size:_102400/known_length/io.ReadAll-10                  	    1000	     44423 ns/op	  514401 B/op	      24 allocs/op
BenchmarkReadAll/size:_102400/known_length/bytes.Buffer+io.Copy-10        	    1000	      9229 ns/op	  106592 B/op	       7 allocs/op

BenchmarkReadAll/size:_1048576/unknown_length/io.ReadAll-10               	    1000	    474003 ns/op	 5241202 B/op	      33 allocs/op
BenchmarkReadAll/size:_1048576/unknown_length/bytes.Buffer+io.Copy-10     	    1000	     81912 ns/op	 1048722 B/op	       8 allocs/op

BenchmarkReadAll/size:_1048576/known_length/io.ReadAll-10                 	    1000	    457693 ns/op	 5241196 B/op	      33 allocs/op
BenchmarkReadAll/size:_1048576/known_length/bytes.Buffer+io.Copy-10       	    1000	     80061 ns/op	 1048674 B/op	       7 allocs/op
PASS
ok  	github.com/elastic/elastic-agent-libs/transport/httpcommon	1.543s

Related issues

• Relates Replace the ioutils.ReadAll usage beats#36151

[elasticmachine]

💚 Build Succeeded

• Buildkite Build
• Commit: 212ef1b

cc @rdner

[amitkanfer]

❤️

[cmacknz]

golang/go#50774 appears to be the upstream Go issue tracking the inefficiency of io.ReadAll here. Linking it for future reference.

Quoting the explanation of what causes this:

The underlying reason is that ReadAll() uses append to resize its buffer, which goes up by a factor of 1.25 (slightly more in latest Go, but asymptotically the same) each time.

One might measure these alternatives on various axes:

number of bytes allocated and freed
max size of heap required to read a given size
amount of copying while resizing
percentage waste in returned buffer

The current implementation is better on the last one, and much worse on all the others.

append is designed to minimize space wasted at the end of the buffer