Update boxcar

[ibraheemdev]

It looks like the reason for the regression in #688 might have been simpler than I thought and come down to some poor codegen when allocating new boxcar buckets (see ibraheemdev/boxcar#28). Hoping CodSpeed also picks up these changes.

[netlify]

✅ Deploy Preview for salsa-rs canceled.

Name	Link
🔨 Latest commit	1c301d2
🔍 Latest deploy log	https://app.netlify.com/sites/salsa-rs/deploys/67dc1958657cd20008dc2b1e

[codspeed-hq]

CodSpeed Performance Report

Merging #696 will improve performances by 3.09%

_{Comparing ibraheemdev:boxcar-alloc-zeroed (1c301d2) with master (7fe054e)}

Summary

⚡ 1 improvements
✅ 11 untouched benchmarks

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	new[InternedInput]	5.8 µs	5.7 µs	+3.09%

[ibraheemdev]

The CodSpeed regression is interesting.. I see a very clear improvement locally.

[davidbarsky]

Maybe one way to handle the noise from codespeed is to look at instruction counts before/after the change?

[ibraheemdev]

Looks like this is a win across the board now. I suspect that's due to #725. I'm not sure what allocator CodSpeed was using previously, but the change should be a clear improvement.