Add a benchmark for report_json parsing

[Swatinem]

This currently yields the following numbers for me locally:

pyreport           fastest       │ slowest       │ median        │ mean          │ samples │ iters
├─ complex_report  452.4 ms      │ 489.2 ms      │ 465 ms        │ 466.9 ms      │ 10      │ 10
╰─ simple_report   17.27 µs      │ 224 µs        │ 18.12 µs      │ 19.79 µs      │ 25076   │ 25076

[codecov]

The author of this PR, Swatinem, is not an activated member of this organization on Codecov.
Please activate this user on Codecov to display this PR comment.
Coverage data is still being uploaded to Codecov.io for purposes of overall coverage calculations.
Please don't hesitate to email us at [email protected] with any questions.

[matt-codecov]

this looks much better than my "just run it 3 times with /usr/bin/time -plh and eyeball it" approach lol

looks like divan has a custom allocator for profiling memory usage https://docs.rs/divan/0.1.14/divan/struct.AllocProfiler.html which is nice for us. those should maybe be separate benchmarks though since i think instrumenting the allocator can destroy runtime?

thank you for setting this up for us! this will be great

[matt-codecov]

is it enough to have this in bench or do we want this in release too?

[Swatinem]

release would sure be nice, that way any panic that might be caught on the PyO3 layer gets a proper stack trace as well.

[Swatinem]

I can sure give divans AllocProfiler a shot, thus far when I tried it, its output was rather messy. But maybe the reason was that I was rather benchmarking stuff that had hardly any allocations at all.