Optimize `binary_decode/1` for lolspeed
binary_decode/1 to a fully binary implementation (instead of being lazy and just leveraging
The current implementation of
binary_decode/1 benchmarks about 20~30% slower than
decode/1 due to reading in a list and then converting to a binary for string objects.
binary_decode/1 should be almost as fast as
decode/1 in a native binary implementation (or at least as fast as JSX).