Binary search

For large data it may be worth it to implement binary search. For write it's simply a matter of moving the split point. For write we need to change the search algorithm to split at byte boundaries between some specific positions and do continuous writing outside of those. Those two positions are boundaries for the binary search algorithm.

This is not currently needed by me but if anyone is interested seriously or for fun feel free to send a PR.

Start by reading the crate's existing std::io::Read checking implementation; the issue does not name files or tests. Work out the byte-boundary split positions and how binary search would affect reads and writes, then verify behavior on large data before considering the change complete.

Written by the indexing model from the issue text.