SERVER-117635 added support to the cbr-for-no-mp-results strategy for passing the initial multiplanner's MP stage to get_executor. To ensure the plan was still cached with the correct number of works, it changed the strategy to complete the plan's remaining number of works to what would be expected without this resumption logic.

This was made even if the caching of the plan is not needed, which caused the optimization time to increase for synthetic benchmarks (which run with the plan cache disabled).

Restrict the lines linked above with a condition similar to get_executor's shouldMultiPlanForSingleSolution.