Force authoring on first slot after runtime upgrade

[JoshOrndorff]

Currently the consensus worker predicts eligibility on a slot-by-slot basis. There is also an option to skip the prediction and force authoring. Typically nodes will not want to force authoring because it makes their logs noisy and wastes CPU time, however it can be a nice tool to have when diagnosing slot prediction issues.

Because of paritytech/polkadot-sdk#64 any storage migrations related to the process of mapping consensus ids to runtime ids can cause prediction problems. While it is hoped that the upstream issue will be fixed, in the meantime, it is useful to have the client-side force authoring in the first slot after a runtime upgrade.

This was first developed when nimbus still shared a repo with cumulus. See https://github.com/purestake/cumulus/commits/elois-force-authoring-when-runtime-change. However this change was not yet brought back into the main nimbus branch.

If it becomes necessary to use this workaround again, I prefer to pass an enum to the client-side worker:

pub enum SkipPrediction {
  Always,
  Sometimes,
  AfterRuntimeUpgade,
}