Replace peak with more efficient peek

[davidhewitt]

Attempts to make converting peak to u8 zero cost.

WIP need to correct the numerical branches to handle invalid bytes (because that's the only place where peak may now error).

[codspeed-hq]

CodSpeed Performance Report

Merging #48 will improve performances by 13.95%

_{Comparing dh/optimize-peak (530ce06) with main (36b2354)}

Summary

⚡ 2 improvements
✅ 35 untouched benchmarks

Benchmarks breakdown

	Benchmark	main	dh/optimize-peak	Change
⚡	pass2_jiter_iter	22.6 µs	19.8 µs	+13.95%
⚡	true_array_jiter_iter	28.4 µs	25.7 µs	+10.68%

[codecov]

Codecov Report

Merging #48 (530ce06) into main (36b2354) will increase coverage by 3.54%.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #48      +/-   ##
==========================================
+ Coverage   89.56%   93.11%   +3.54%     
==========================================
  Files           8        8              
  Lines        1045     1046       +1     
==========================================
+ Hits          936      974      +38     
+ Misses        109       72      -37     

Files	Coverage Δ
src/jiter.rs	99.46% <100.00%> (+16.04%)	⬆️
src/parse.rs	98.74% <100.00%> (+0.58%)	⬆️
src/python.rs	94.28% <100.00%> (+0.05%)	⬆️
src/value.rs	100.00% <100.00%> (ø)

... and 1 file with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 36b2354...530ce06. Read the comment docs.

[davidhewitt]

Will there's definitely something here but it has some error case no longer matching serde.

That's not a huge surprise because the way this changes Peak is to assume that it contains the start of a value, and when number parsing fails only then to backtrack and throw the "expected value" error. So one or more number parsing branches must not be backing out correctly.

[samuelcolvin]

LGTM, can we cover those cases that codecov is complaining about?

[davidhewitt]

I decided to close #50 at the same time.