[FEAT] Profiling: array size

[samnlindsay]

Is your proposal related to a problem?

When working with array columns it is useful to be aware of the distribution of number of elements in those arrays. If they're almost always single-valued, maybe worth reconsidering whether an array is necessary. If they can get too large, this should be flagged before developing a model where these large arrays could be problematic.

Describe the solution you'd like

Prototype chart generated in https://github.com/moj-analytical-services/data_linking/pull/795

Describe alternatives you've considered

Additional context

Related to #1064 / #1397 - do we profile array contents at the moment? If not, a combined array profiling chart would be useful (i.e. size distribution + value distribution for each array column)