Consider to drop mimalloc

[gi0baro]

Given

• Replace MiMalloc w/ default allocator pydantic/pydantic-core#900
• fix: mimalloc virtual memory gi0baro/geohashr#3

it seems mimalloc causes large virtual memory allocations.

Some quick tests on CPython 3.10 on Apple Silicon shows a few percentage points performance degradation (RPS) when switching to default allocator.

Before a drop/replacement action I'd like to make more tests on memory usage and performance degradation, gonna update this in the process.

[Aeron]

My two cents, if I may:

• Mimalloc is still a viable solution for MUSL-based platforms since the default allocator there is not very efficient;
• Also, CPython kind of moves towards Mimalloc potentially (gh-90815: Add mimalloc memory allocator python/cpython#31164).

The latter might be irrelevant to any decision here, just a curious fact.

[gi0baro]

• Mimalloc is still a viable solution for MUSL-based platforms since the default allocator there is not very efficient;

We can definitely use mimalloc just on MUSL builds.
Anyways, this is just a reminder for myself, before taking any decision I gonna update this with benchmarks and profiling results :)