ReshardingCoordinator::installCoordinatorDoc should update in memory state before calling logChange

logChange will eventually call shard::runCommand, which has checks if operation context was interrupted. This is problematic because installCoordinatorDoc is usually used in conjunction with a corresponding write to update coordinate state. And if it gets interrupted while calling logChange, the in-memory state will remain in the pre-update state and will incorrectly retry ops if inside a WithAutomaticRetry.