Fix misaligned pointer dereference when storing UserData

[rlidwka]

see discussion in #210

I've also added a test, which verifies userdata storing/retrieving with various types, and may also reproduce a problem with older code. You can run it with the following command:

cargo test --package physx --lib -- traits::user_data::tests --nocapture

line 28 is where the bug is at, lines 24-25 are changed only for consistency reasons

[tgolsson]

Awesome work, thanks! Quick cargo fmt and we should be good to go!

[rlidwka]

Quick cargo fmt and we should be good to go!

Okay, I've added formatting changes as the second commit.

Note: cargo fmt errors happened because rust people finally figured out "the only right way" to write let..else statements in a recent rust release, and those changes have nothing to do with the changes in the first commit.

[tgolsson]

👍 Yeah; figures. Thanks for fixing anyways. I'll see if I can cut a release later this week, a bit swamped right now.

[rlidwka]

I've just noticed that other two methods in the same file use double-unsafe blocks, i.e.:

unsafe fn fn get_user_data_mut(...) {
    unsafe {
        ...
    }
}

Example:

physx-rs/physx/src/traits/user_data.rs

Lines 51 to 62 in bbfa160

	unsafe fn get_user_data_mut(this: &mut Self) -> &mut Self::UserData {
	unsafe {
	if size_of::<Self::UserData>() > size_of::<*mut c_void>() {
	// Data is stored in a Box<UserData> on the heap, and userData is just a pointer to it.
	&mut *((*this.user_data_ptr_mut()) as *mut Self::UserData)
	} else {
	// DATA_SIZE <= VOID_SIZE
	// Data is stored directly in the userData field.
	&mut *(this.user_data_ptr_mut() as *mut *mut c_void as *mut Self::UserData)
	}
	}
	}

What's the reason for that, and should I have wrapped if..else in unsafe block in a similar way?

[tgolsson]

I'm going to guess it comes from this lint: https://doc.rust-lang.org/beta/nightly-rustc/rustc_lint/builtin/static.UNSAFE_OP_IN_UNSAFE_FN.html, though it doesn't seem like we apply it consistently...