v13 migration: Audit and clean up the two-stage cache utilization

Unlike previous migrations, the v13 migration has different behaviour depending on whether or not it's running in "cached" or "uncached" mode. This means that if, for some reason, the cache isn't correctly and completely populated in a previous migration ("pre-migration"), the result will be WRONG.

While this is still correct in the happy path, we need to robust-ify this. In particular, we should consider:
- having some top-level directive about which "mode" to be running in (currently the individual miners and the market attempt to determine that themselves)
- decoupling the miner and market's usages of the cache (currently it would be WRONG if a single miner ran in "un-cached" mode, but the market actor was migrated in cached mode).