No data is persisted during IBD

[JoseSK999]

Previously indexes and headers were saved on the fly (while rn they are kept in a HashMap until calling flush), and flush was called periodically during IBD (now only after validating blocks once synced).

In other words, all the progress is lost if the node crashes during IBD. This may be faster but doesn't seem desirable.

[jaoleal]

Maybe we could offer a Checkpoint option and the user can set how many blocks we can "risk" until they are saved in.

in config.toml

checkpoint = 100_000 # this would make floresta wait every 100_000 blocks until stop everything and save on DB

[Davidson-Souza]

Yeah, this is super annoying. The reason I've stopped flushing and added the hashmap is that kv was killing performance, specially during headers sync. I hope that #251 will give us the possibility of restoring from a shutdown during IBD