op-supervisor: Improve reorg during reset behavior

[sebastianst]

Currently, reorgs that happen while a reset is ongoing may lead to an outdated reset target. This may even more so be the case because resets are now blocking and reorgs may not be executed in the background.

Reorgs and resets are both rare, so we should find a good solution with little complexity to not get to wrong reset targets.

One solution may be to run resets async again, but then there's more complexity, e.g. detecting in other places that a reorg is in progress. Another solution may be to just check that a reset target is valid before sending it to the node, and if not do another reset. This way, if after the reset, and before sending it to the node, a reorg happens, it would be detected.