Implement benchmarking with AirspeedVelocity.jl

[mikeingold]

Changes

• Adds a CI Action (duplicate of this one) that uses AirspeedVelocity.jl to run benchmarks for every new PR
• Adds a benchmark script with a suite containing 14 @benchmark entries, testing integrals of Segment and Sphere (surface), jacobian, and differential.

Closes #21

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 95.75%. Comparing base (6c8cd1f) to head (eedc533).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #119   +/-   ##
=======================================
  Coverage   95.75%   95.75%           
=======================================
  Files          17       17           
  Lines         259      259           
=======================================
  Hits          248      248           
  Misses         11       11           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[mikeingold]

@JoshuaLampert, this seems to work well on my system and should provide us with benchmarks to quantify performance impacts of each PR. The Action script is taken straight from AirspeedVelocity.jl; I don't see any blatant red flags, but you're more knowledgeable in that domain than I am.

[JoshuaLampert]

This looks quite neat! I would like to see the action already running for this PR. Normally GitHub already runs newly added actions already in the PR, where they are added. So let's see if fixing the pull_request_target branch runs the action.

[JoshuaLampert]

It still doesn't run. I'm not sure why.

[JoshuaLampert]

Ahh, there we go. Now it runs, but fails. Let me fix that quickly.

[JoshuaLampert]

There were some old versions.

[github-actions]

Benchmark Results

	main	eedc533...	main/eedc533858dc6f...
Differentials/Differential	0.765 ± 0.066 μs	0.764 ± 0.063 μs	1
Differentials/Jacobian	0.668 ± 0.026 μs	0.683 ± 0.026 μs	0.979
Integrals/Meshes.Segment/Scalar GaussKronrod	4.4 ± 0.22 μs	4.4 ± 0.21 μs	1
Integrals/Meshes.Segment/Scalar GaussLegendre	0.075 ± 0.00051 ms	0.0742 ± 0.00049 ms	1.01
Integrals/Meshes.Segment/Scalar HAdaptiveCubature	7.93 ± 0.034 μs	7.97 ± 0.05 μs	0.996
Integrals/Meshes.Segment/Vector GaussKronrod	6.6 ± 0.075 μs	6.64 ± 0.097 μs	0.994
Integrals/Meshes.Segment/Vector GaussLegendre	0.0881 ± 0.0027 ms	0.0907 ± 0.0044 ms	0.972
Integrals/Meshes.Segment/Vector HAdaptiveCubature	10.9 ± 0.13 μs	10.9 ± 0.14 μs	0.996
Integrals/Meshes.Sphere/Scalar GaussKronrod	0.195 ± 0.00051 ms	0.198 ± 0.00066 ms	0.984
Integrals/Meshes.Sphere/Scalar GaussLegendre	16.8 ± 1.2 ms	16.9 ± 1.2 ms	0.993
Integrals/Meshes.Sphere/Scalar HAdaptiveCubature	0.19 ± 0.0021 ms	0.19 ± 0.0023 ms	1
Integrals/Meshes.Sphere/Vector GaussKronrod	0.237 ms	0.232 ms	1.02
Integrals/Meshes.Sphere/Vector GaussLegendre	18.9 ± 0.46 ms	18.9 ± 0.43 ms	1
Integrals/Meshes.Sphere/Vector HAdaptiveCubature	0.255 ± 0.0057 ms	0.261 ± 0.0089 ms	0.979
time_to_load	1.59 ± 0.0094 s	1.58 ± 0.011 s	1.01

Benchmark Plots

A plot of the benchmark results have been uploaded as an artifact to the workflow run for this PR.
Go to "Actions"->"Benchmark a pull request"->[the most recent run]->"Artifacts" (at the bottom).

[JoshuaLampert]

Everything runs fine now. Also uploading the plots works and updating the comment after a new push works nicely.
(Sorry for all the noise I made in this PR)

[mikeingold]

Thanks for getting this working!