Potential performance improvement

Hi @josephg , is there any reason to allocate new buffer on every encode/decode call?
https://github.com/josephg/fdb-tuple/blob/master/lib/index.ts#L153

The code looks synchronous and in node's single threaded environment, one large (eg 10k) buffer should reduce allocations to near zero and increase performance dramatically. I'm wondering if I might be missing something.