Revert "remove borrow checks from ZalsaLocal"

[ibraheemdev]

This is showing up as a 7% regression on master?

[netlify]

✅ Deploy Preview for salsa-rs canceled.

Name	Link
🔨 Latest commit	b9ae30d
🔍 Latest deploy log	https://app.netlify.com/projects/salsa-rs/deploys/688ce62a72242800084a26d5

[codspeed-hq]

CodSpeed Performance Report

Merging #950 will improve performances by 10.42%

_{Comparing ibraheemdev:ibraheem/revert-unchecked-zalsa-local (b9ae30d) with master (0ca4f4f)}

Summary

⚡ 2 improvements
✅ 10 untouched benchmarks

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	amortized[InternedInput]	2.5 µs	2.2 µs	+10.42%
⚡	new[InternedInput]	4.3 µs	4.1 µs	+4.99%

[Veykril]

That is odd ...

[Veykril]

jeamlloc appears in some diffs, so this might be allocator noise? Though this PR shouldn't be touching things in a way to affect allocations?

Looking at some of the other diffs there they really do not seem related, like dyn_memos getting slower. The only part there that can get slower is the memory access which again points to allocation noise imo.

Can you reproduce this regression on ty? If not I would say we should keep the changes.