fix: fix fault tolerance

[cchudant]

Pull Request type

• Bugfix

What is the current behavior?

DB can get corrupted

What is the new behavior?

the mix of these:

• disable wal
• enable atomic flush
• disable auto flush (apprently you can't really disable it completely, or at least I havent found how -- once the write buffer is filled rocksdb has to flush the memtables. hopefully the write is big enough for the later blocks)
• flush at the end of a block if 5 sec has passed since last flush

enable us to be fault tolerant.

We can't just use transactions as they are not multithreaded and we really need parallelism for perf

Does this introduce a breaking change?

I don't think so

Other information

Added a little bit of cleanup