[Dump] Catchup block 0

[gregdhill]

With a dump-restore, previous events are stored in block zero which is an issue when catching up a node as Tendermint will not replay this block, resulting in an appHash collision.

[silasdavis]

The idea with restore is that we explicitly do not pay for consensus on transactions committed in a previous chain, so Tendermint not gossiping those is as-designed.

Effectively on a restore chain genesis == genesis doc + dump state. They should be distributed together