chore: improve /healthz endpoint performance

[inkel]

👋 team. First and foremost: thank you for open sourcing and maintaining this project, it's been really useful. Today while working on an unrelated issue I started to investigate the /healthz endpoint in Atlantis and found that it could be improved in both speed and memory allocations.

The first thing I did was adding a benchmark to have a baseline for comparing my changes:

goos: darwin
goarch: amd64
pkg: github.com/runatlantis/atlantis/server
cpu: Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz
BenchmarkHealthz-12       100000               531.0 ns/op           112 B/op          4 allocs/op
PASS
ok      github.com/runatlantis/atlantis/server  0.370s

Then I implemented my improvements. I noticed that the response body is always the following:

{
  "status": "ok"
}

However, I noticed the body was built by defining an inline struct and calling json.MarshalIndent on it, checking if there was a marshaling error which should never happen, so I removed all of that code and simply define a variable with the expected body. The benchmarks results already show improvements:

goos: darwin
goarch: amd64
pkg: github.com/runatlantis/atlantis/server
cpu: Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz
BenchmarkHealthz-12       100000                94.31 ns/op           40 B/op          2 allocs/op
PASS
ok      github.com/runatlantis/atlantis/server  1.431s

Lastly, I realized there was no need to always initialize this variable, so I moved it to a private package variable and ran the benchmarks again:

goos: darwin
goarch: amd64
pkg: github.com/runatlantis/atlantis/server
cpu: Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz
BenchmarkHealthz-12       100000                60.20 ns/op           16 B/op          1 allocs/op
PASS
ok      github.com/runatlantis/atlantis/server  0.336s

Most of all of the allocations were removed and the speedup is ~90%! In all steps I actually generated more benchmarks results using the following command:

go test -v -{run,bench}=Healthz -benchmem -benchtime=100000x -count=100 ./server/

With this I was able to later run benchstat to compare the results:

name        old time/op    new time/op    delta
Healthz-12     444ns ± 7%      52ns ±25%  -88.33%  (p=0.000 n=84+100)

name        old alloc/op   new alloc/op   delta
Healthz-12      112B ± 0%       16B ± 0%  -85.71%  (p=0.000 n=100+100)

name        old allocs/op  new allocs/op  delta
Healthz-12      4.00 ± 0%      1.00 ± 0%  -75.00%  (p=0.000 n=100+100)

Anyway, I hope you find this useful and worthy of contribution! 😸