[VIndex] Need a faster implementation of disk-based storage #64
Description
The current implementation of the verifiable index doesn't scale to CT-size logs. Or rather, the number of keys that indexing CT generates.
The in-memory version requires too much RAM, and the disk-based version isn't fast enough to process this many entries in a reasonable timeframe.
There is a faster version in progress, and we should incorporate that eventually.
In the meantime, we need some way to cut down CT indexing to a reasonable size. Options:
- MapFn that outputs only a deterministic subset of entries (e.g. only those for certain domains, TLDs, or some other predicate)
- Create a local log with a reasonable number of entries populated from a prod log