SIP: Resume Replay

[mvandeberg]

If steemd is terminated during a replay, it likely shuts down in the middle of processing a block, making it impossible to resume the replay. And if it does shut down cleanly, it will continue by syncing from the live network despite having the blocks locally.

We need some extra infrastructure to ensure that if steemd is shut down mid reindex, it shuts down completely.

Then, if a node starts back up and the state head block is earlier than the block log head block, the node should enter a "resumed" replay state that takes blocks from the block log rather than syncing those same blocks from the p2p network.

This will greatly increase the robustness of steemd nodes by providing some much needed "quality of life" features to node operators.

Combined with the platform independent state files (#3084) this will make it trivial to generate intermediate snapshots, and will reduce the headache incurred by stopping a replay close to the head block.

[fernandosoder]

Please, do it.