rr chaos mode for stress testing garbage collection

x-post from https://github.com/JuliaPy/PythonCall.jl/pull/520#issuecomment-2241796411. Probably useful for when https://github.com/JuliaPy/PythonCall.jl/pull/529 lands.

Regarding testing, it could be nice to try something like [rr's chaos mode](https://robert.ocallahan.org/2016/02/introducing-rr-chaos-mode.html) to assault the garbage collector in multithreaded mode and reveal any potential thread safety issues.

It is tricky to start Julia with rr from within Python – see https://github.com/JuliaLang/julia/issues/52689. So the test would have to be run from the Julia side of PythonCall.jl, and open Python from there, rather than the other direction. But maybe that would be good enough for testing this.

I believe the Julia version of this can be done with

```julia
julia --bug-report=rr-local,chaos
```

which starts a new Julia process using BugReporting.jl.

BugReporting.jl looks to launch `rr` using

https://github.com/JuliaLang/BugReporting.jl/blob/fec2ece8be317462fe8b93ffca56e5b228d0b341/src/BugReporting.jl#L162

```
rr_cmd = `$(rr()) record $rr_flags $julia_cmd $julia_args`
```

where `rr_flags` contains `--chaos` for chaos mode.

So perhaps the BugReporting.jl package could be installed for testing, which automatically installs `rr`. This could be used for referencing the `rr` executable with `BugReporting.rr()`.

cc @EricPhanson

Start by reading the Julia-side testing approach described in the issue and BugReporting.jl's rr invocation at the linked source. Try `julia --bug-report=rr-local,chaos` with PythonCall launched from Julia. Done means there is a repeatable chaos-mode stress test that exercises PythonCall's multithreaded garbage-collection path.

Written by the indexing model from the issue text.