op-node: reset engine through events

[protolambda]

Description

This changes the op-node to reset the engine by:

• detecting the reset error, and then emitting a request to the engine-resetter to compute where to reset to
• once computed, the reset is forced to the engine.
• once completed, the reset is confirmed to the derivation pipeline

The old ResetEngine function was split as follows:

• the computational part, basically just sync.FindSyncStart call: happens on request event
• the application of the start point to the engine: happens on force event
• the safe-head notifications: op-node driver handles it for now, when receiving the confirmation event
• reset is confirmed to pipeline

Note that events are still processed synchronously, so the code won't spawn more reset-error events from the pipeline until the reset has been confirmed. Once we make this async we'll need to hold a lock somewhere that a reset is in progress.

Now it can be composed nicely with the other event derivers.

Tests

The reset code-paths are used in almost every test, and no branching functionality is added, so no new tests.

Metadata

Fix #10958

[ajsutton]

LGTM.