GitHub Action to Comparatively Benchmark PRs

[blue-jaye-121]

Description Of Changes

With the introduction of benchmarking into MetPy, the next step is to incorporate a GitHub action that will benchmark PR's into MetPy's main branch alongside the current performance to make sure merges will not negatively impact the performance of MetPy. To do so, I've added .github/workflows/benchmark-pr.yml, which is a GitHub workflow that uses asv to comparatively benchmark PR's that are labeled with the label benchmark and fails if any benchmark is 10% or more slower than the current main branch's performance. This check takes ~10 minutes, which is why only labeled PRs will be benchmarked so that those that do not change the calc module can be merged more quickly.
In order to test the action, I used my fork https://github.com/blue-jaye-121/MetPy and opened two pull requests onto this feature branch. The first, blue-jaye-121#3, is not expected to fail. This branch is identical to the feature branch except for a new fakechange.txt file which is empty. As you can see from the results,

By contrast, blue-jaye-121#4 is identical to the feature branch except for a sleep function on the scale_height() function. As you can see, this check fails:


since the scale_height() function is over 600% slower.

Checklist

• Fully documented