Use a deque for the ring buffer for better dequeue performance

[mlw]

Replaces the underlying type of the ring buffer to be a std::deque. Previously this was a std::vector. While we had reserved space for the vector which would prevent growth and make Enqueue always cheap. Dequeue ops would still require shifting the items in the vector. std::deque gets rid of this limitation.

There is a tradeoff in memory usage here as std::deque will use more. But current use cases only require small ring buffers. If we ever need larger ring buffers, we should move to a more robust ring buffer implementation that manages it's own contiguous memory.