Pop performance (addresses #15)

[peteraisher]

I've managed to get a decent speed-up for pop() without an extensive rewrite.

Profiling testPopPerformance() shows that ~ 70% of the execution time for pop() is taken up by MutableCollection.swapAt(_:_:) and another ~ 13% is Array.subscript.getter.

Implementing the swap and the sink inside a call to Array.withUnsafeMutableBufferPointer(_:) gives a speedup of around 9x (on the release build) on my machine and puts the performance of pop() on a level with push(_:) and remove(_:).

I think the performance gap must be a case of the compiler just not figuring out the optimization for pop(), since all of the other public functions which call sink(_:) and swim(_:) are plenty fast.

[davecom]

Thank you, @peteraisher. This is the best of both worlds—code that I can maintain and that addresses the issue. I really want to thank you for the care and consideration you put into it. I and all the other users of SwiftPriorityQueue thank you.