overlaps for Data.Vector.Mutable behaves oddly for empty vectors

The `basicOverlaps` implementation for Data.Vector.Mutable is:

``` haskell
basicOverlaps (MVector i m arr1) (MVector j n arr2)
= sameMutableArray arr1 arr2
&& (between i j (j+n) || between j i (i+m))
where
between x y z = x >= y && x < z
```

This will return `True` for two vectors where two vectors share the same mutable array and offset, but one has a zero length and the other has a non-zero length. I cannot find a precise definition of what it means for possibly-empty vectors to overlap, so it's hard to say for sure this is wrong, but it certainly disagrees with my understanding of what `overlaps` ought to mean.

I suggest this implementation instead:

``` haskell
basicOverlaps (MVector i m arr1) (MVector j n arr2)
= sameMutableArray arr1 arr2 && i < j + n && j < i + m
```

That will still report an overlap for an empty vector with an offset in the middle of a containing non-empty vector, though. It's even less clear what the right answer is there. So I guess what we really need is a precise specification first, and then we can write an implementation that matches it.