Warning or error for inefficient use of fill missing points routines

# ✨ Feature Request
The methods for filling missing points can take a long time when used to fill many, many missing data points (e.g. filling a continent rather than a coastline). In these circumstances, a floodfill or similar is likely to be a better approach to fill the bulk of the missing data.

Can we raise a warning or an error when the number of points being filled exceeds a threshold (e.g. if there's more than 1% missing data)?

Note that the KDTree and spiral search approaches would likely have different thresholds - so the threshold should be specific to the algorithm being used.

Start by locating the missing-point filling routines and their KDTree and spiral-search implementations. Determine how each algorithm measures points being filled and where an algorithm-specific threshold can be applied; done means excessive filling produces the agreed warning or error while ordinary fills remain unaffected.

Written by the indexing model from the issue text.