Use volatile to prevent the compiler from optimizing alias pointers

[jjyr]

Fix #16

I suspect the compiler may have discarded line 44 due to optimizations, compiler ignoring the potential for pointer aliasing. I haven't confirmed from the disassembly, but this PR fixed the issue.

buddy-alloc/src/fast_alloc.rs

Lines 44 to 45 in 1a3e2e2

	(*(*list).next).prev = p;
	(*list).next = p;

I have also modified the code of buddy alloc to prevent potentially similiar errors.

After the fix, the performance even improved 8% ~ 10% on my macbook, it is because we removed a write_unaligned.