`mi_malloc_satisfies_alignment` returns false for small alignments

`mi_malloc_satisfies_alignment` seems like it will not do the right thing for a allocation request with a explicit-but-small alignment request. This can lead to some functions calling mi_malloc_aligned over mi_malloc, which will hurt performance.

As a concrete example, the function returns false for an alignment of 1, but any pointer would satisfy this trivially. (I suspect in reality the first check should be `alignment <= sizeof(void*)`?)

https://github.com/microsoft/mimalloc/blob/13a4030619edada133e1a3c06c6287ad5a9c74c1/include/mimalloc-internal.h#L237

On a related note, what *is* the test for a given `(size, alignment)` that I should go through the unaligned function? I'm attempting to improve Rust's mimalloc integration, and for Rust, every allocation request comes with both size and alignment. I had been assuming `alignment <= alignof(maxalign_t) && alignment <= size` is the test, but then I noticed this.

Start at include/mimalloc-internal.h around line 237 and inspect mi_malloc_satisfies_alignment, mi_malloc, and mi_malloc_aligned. Check concrete small-alignment cases such as alignment 1 and compare the function's result with the proposed pointer-size boundary. Done means the alignment decision is correct and the (size, alignment) rule for using the unaligned path is clear.

Written by the indexing model from the issue text.