NockVM persistent memory via LMDB

Edward vaguely recalls a reason he didn't think this would work, but we should find out for ourselves and see how hard it would be to fix.

The idea with the LMDB implementation is that it's closer to a true PMA because we'd be working with a direct and compatible in-memory noun representation via LMDB's mmap. It is highly likely that we'll need to go through multiple iterations of this.

This needs #7

The issue names no files, tests, or entry points; begin by investigating the proposed LMDB-backed persistent memory approach and the dependency on issue #7. Done means determining whether the approach works, documenting the required iterations, and addressing the identified obstacles.

Written by the indexing model from the issue text.