Contribute stuff we need to `rocksdbpy`

This would be

• fast key count estimate,
• choosing compaction settings,
• and many other missing (set_parameter_value methods),
• maybe options like "bloom filters" to speed up reads (something that tells you if something is probably in the database or not, based statistics).

rocksdbpy is very old and seems not really maintained.

We should ask in the Tweag GenAI group or in @arobert's department whether we can spend time of this.