Draft: Batched indexing for files bigger than RAM

[mkrd]

The goal here is to be able to index a file that is bigger than the available RAM.

This is not good enough yet because it assumes that a key is at most 10 million characters long