Currently we measure planning and execution time per query shape in the metrics:

metrics.totalExecMicros.[min|max|sum|sumOfSquares]
metrics.queryPlanner.planningTimeMicros.[min|max|sum|sumOfSquares]

Introduce additional metrics for different plan selection strategies.

This can be similar to the queryLatencies metrics in serverStatus(): multiPlanner, costBased, singlePlan, cachedPlan, express. For each planning strategy add its count and exec and planning time aggregated statistics. For instance:

metrics.queryPlanner.costBased.count
metrics.queryPlanner.costBased.planningTimeMicros.[min|max|sum|sumOfSquares]
metrics.costBased.totalExecMicros.[min|max|sum|sumOfSquares]

Adding histograms instead of aggregated statistics is expected to be too high overhead.

If we need to reduce the number of new metrics, as a minimum we should introduce the metrics for the costBased plan ranker.