Upgrade maturin to 0.14

[messense]

Cargo.lock is now included in sdist by default.

[codspeed-hq]

CodSpeed Performance Report

Merging #336 maturin-0.14 (1e8c478) will not alter performances.

Summary

🔥 2 improvements
❌ 0 regressions
✅ 61 untouched benchmarks

🆕 0 new benchmarks
⁉️ 0 dropped benchmarks

Benchmarks breakdown

	Benchmark	main	maturin-0.14	Change
🔥	test_many_models_core_model	100.4 ms	15.9 ms	84.11%
🔥	test_list_of_nullable_core	1,019 µs	905 µs	11.19%

[art049]

Do you think this performance change is accurate? @messense @samuelcolvin
This seems a bit extreme

[messense]

maturin upgrade shouldn't affect performance in general.

[samuelcolvin]

test_many_models_core_model matches the change on #327, so I would guess the strange result is on main.

But generally I do see these wide variations from time to time, @art049 any idea why?

[art049]

There are clearly 2 results clusters:

Between the 2, there are 5 times more instructions executed in the slowest one and 4~ times more memory writes.
I think it's probably on the Python side, maybe the garbage collection. I'll try to run some tests this afternoon.

Do you think it could also be something coming from the Rust part? This seems unlikely to me but I'm not familiar yet with the internals of Rust.
As you mentioned @samuelcolvin, I'll try to make the hashmap use a fixed seed to see if this is the root cause but with such a big difference I don't think it's the root cause of the flakiness.

Did you see something like this for this benchmark in the results from pytest-benchmark?

[art049]

I managed to reproduce:

In orange with garbage collection not disabled and in yellowish green with GC disabled. I'll disable it by default when running tests the pytest plugin.

[samuelcolvin]

thanks so much @messense.