RocksDB as WAL persistence, separate from nockvm in-memory

I think we'd prefer it if #8 worked out, but if it doesn't, this takes a different tack and focuses less on PMA and more on WAL. The idea is that we'd be using RocksDB as an embedded database akin to SQLite but it's much better suited to high write throughput of binary-tree structured values like Nouns which is the universal data structure in Urbit's Vere and NockVM.

Start by reviewing issue #8 and the repository context around WAL, PMA, and nockvm. The issue does not name files, tests, or a concrete acceptance criterion; before implementation, clarify whether this RocksDB-based persistence direction is still wanted and define what completion would mean.

Written by the indexing model from the issue text.