Evaluate an idea about PGO for optimizing heap allocations with Mimalloc

Hi!

A few days ago, I was watching a talk about TCMalloc's work on trying to optimize heap allocations based on runtime gathered statistics (like PGO). The materials are located [here](https://github.com/valexey/tcmalloc_hot_cold/tree/main) (slides are [here](https://github.com/valexey/tcmalloc_hot_cold/blob/main/TCMalloc%20hot%26cold.pdf)). I recommend to read all slides but the PGO part starts from page 36.

Since `mimalloc` is a performance-oriented allocator, I wanted to suggest an idea to `mimalloc` authors to evaluate a similar idea for `mimalloc`. I don't have enough knowledge about `mimalloc`'s internals, but I guess if the idea kinda works/promising for TCMalloc - it could be useful for other allocators too.

Thank you.

Start by reading the linked TCMalloc hot/cold materials, especially the PGO section beginning on page 36, then study mimalloc's allocator internals. The issue is complete when the PGO-inspired heap-allocation optimization has been evaluated for mimalloc and its usefulness or limitations are documented.

Written by the indexing model from the issue text.