Add test to limit allocations of Primer::Classify

[BlakeWilliams]

Reduce allocations from 97 to 38 when using a "basic" Primer::Classify#call implementation.

closes part of #155

[vercel]

This pull request is being automatically deployed with Vercel (learn more).
To see the status of your deployment, click below or on the icon next to each commit.

🔍 Inspect: https://vercel.com/primer/view-components/eakj5jko7
✅ Preview: https://view-components-git-bmw-allocations.primer.vercel.app

[BlakeWilliams]

I ran some benchmarks testing the new classify vs the classify on main and it looks like we're ~1.6x faster too!

primer_view_components λ bundle exec rake benchmark
/Users/blake/.rbenv/versions/2.7.1/bin/ruby ./performance/benchmark.rb
Warming up --------------------------------------
            classify     5.291k i/100ms
        old classify     3.188k i/100ms
Calculating -------------------------------------
            classify     51.519k (± 6.5%) i/s -    259.259k in   5.053442s
        old classify     32.399k (± 4.7%) i/s -    162.588k in   5.029117s

Comparison:
            classify:    51519.4 i/s
        old classify:    32399.3 i/s - 1.59x  (± 0.00) slower

[BlakeWilliams]

I kept these names the same as from before the refactor, but happy to update them to be more descriptive.

[joelhawksley]

Where'd this value of 4 come from? Does this value vary a little in local test vs. CI?

[BlakeWilliams]

I should probably add a comment, but I believe the difference is between Ruby and Rails versions, since we run a matrix of versions in CI.

The other thought was to make a build specifically for testing allocations against 2.7, but I thought I'd submit this as-is and get thoughts/feedback first.

[BlakeWilliams]

Updated the test to include all, or most of the keys we support. Running the test against main resulted in 204 allocations, this branch resulted in 87 allocations.