Use a memory pool for binary reader string decoding buffers

[tgregg]

...thereby reducing the cost of repetitive reader instantiation. See #369 for a similar change made to the binary writer's encoding buffers.

[tgregg]

I have a solution that works with the incremental binary reader proposed in #355. I'll wait until #355 is merged to create a PR.

Subsequently, I will create a similar PR for the existing non-incremental binary reader, which should be able to reuse the same pool used by incremental readers.