Duplication of recipes to produce multiple types of plot for single diagnostic

[jwarner8]

There was a discussion at the CSET day in Jan about the fact that some diagnostics, like the age of air, could be visualised in different ways after being computed. For example, at the moment, spatial maps are produced, but it could also be useful to plot a domain average time series. To avoid interlinking dependencies, the current preferred approach is to have two separate recipes that produce each plot, and thus duplication of compute to create the diagnostic. This might become problematic if diagnostics are computationally expensive or scale up poorly. Opening issue to capture this discussion topic.