SERVER-72997 made it so that the maxdiff algorithm could produce valid histograms where the number of buckets is the number of types + 1 (or equal to in the case the first type/bucket having a single value). As pointed out in review, we should include a guardrail for user invocations of the analyze command which asserts the given number of buckets is large enough, before we begin constructing the histogram.

This likely needs to happen in the implementation of the generateStats accumulator.