Proposal to Merge `mmapmem` into `mapallmem`: Need Clarification on `data == 2`
I'm planning to merge `mmapmem` into `mapallmem`  (see [commit 93ebb0c](https://github.com/devarajabc/box64/commit/93ebb0c58a434bb6c0f39503b88a8b9ee4047139)), by using the `data` field to indicate whether a memory fragment is allocated via `mmap`. Since all memory mapped in `mmapmem` is also mapped in `mapallmem`, maintaining both trees may cause memory redundancy.

In the proposed scheme:

- `data == 1`: Memory fragment is allocated.
- `data == 2`: Memory fragment is allocated by `mmap`.

However, I noticed that the `data` field is already being assigned the value `2` in the `reverveHigMem32()` function
```c
//src/custommem.c
void reverveHigMem32(void) {
    // -- skipped --
    rb_set(mapallmem, (uintptr_t)cur, (uintptr_t)cur + cur_size, 2);
    // -- skipped --
}
```
I couldn't find any documentation or code that clearly distinguishes this usage of `data == 2` in `mapallmem.`

So my question is: what does `data == 2` represent in this context?
Alternatively, would it be acceptable if I use `data == 3` in `mapallmem` to indicate memory fragments allocated by `mmap`?

To test this approach, I ran  `chess.exe` with `box64` ([commit 93ebb0c](https://github.com/devarajabc/box64/commit/93ebb0c58a434bb6c0f39503b88a8b9ee4047139)) and `wine`  (same config as in [issue #2511](https://github.com/ptitSeb/box64/issues/2511)). 
Below is the result after merging `mmapmem` into `mapallmem`:

![Image](https://github.com/user-attachments/assets/1584c8e6-0af5-451b-8c90-0fdbbb5403bb)
Compare to Box64 (v0.3.5 73dfe869):
![Image](https://github.com/user-attachments/assets/f50513fe-e4f9-453a-9478-76d8981facdc)

As shown above, the total number of red-black tree nodes was reduced by approximately 81.
Since each node occupies 56 bytes, this translates to a memory saving of roughly 4.4 KB, which can become more significant when running larger programs with extensive memory mappings.