Fix memory breakpoint when checking the middle of the data

[aldelaro5]

Instead of just repeating what the commit says, here's a more practical example so it's easier to understand what the issue is and how I fixed it.

In Super Mario Sunshine, the RNG seed address is 0x8040CE30 and is a word (4 bytes long). Currently, in master, if you were to add a memory breakpoint that would break on this address, the address MUST be included in the range or must EXACTLY be this address in the case of a single address memory breakpoint. The problem comes from the fact that it didn't JUST wrote to this address, it also wrote to the next 3 one because it is writting 4 bytes so a single address memory breakpoint to 0x8040CE33 for instance would not break, even though it did wrote/read that address, it just didn't STARTED the read/write there.

So I changed how GetMemCheck works, I added a size parameter so it can adapt itself depending on how much you are writting (by default 1 because only the MMU memcheck function needs to pass an exact size, the rest actually only care about memchecks being present which is the case for the GUI display). Then, after a lot of tinkering around to get the conditions right, I actually found out that you don't even need to do something different whether the memory breakpoint is ranged or not. The reason the function can be so simpler is because in the case of non ranged memory breakpoint, the start address is equal to the end one so the same condition can apply to everything.

Because of how minimal this function does, it doesn't impact performance, even though it's called very often. I actually wasn't able to notice any performance dips on rs2 and the intro of rs3.

Last note, to test this on Windows, you need to apply #5004 otherwise they just won't work.

[lioncash]

sizes should be represented with size_t.

[degasus]

I'd prefer u32 instead of size_t. We're talking about memory sizes of hte GC, not of the host.

[lioncash]

@degasus This will also make it a requirement to cast every sizeof expression, to avoid type truncation. This is why most of the copy routine use size_t. That's kind of a bad idea, in my opinion, especially given those are the most appropriate inputs most of the time. It would likely be better to add an assert or check against the value range, if it's that important.

[aldelaro5]

Changed everything to size_t because I am a bit confused tbh.

EDIT: decided to stay with size_t because as @lioncash said, it reduces problem with sizeof and it's still a size, you don't loose that much meaning by not using u32.