`scan_data()` is very slow on large datasets

[hadley]

Specifically the interactions and correlations sections:

pointblank::scan_data(ggplot2::diamonds, sections = "IC")

Iteractions takes ~30s which isn't too surprising, but correlations takes a very long time, which does surprise me.

[rich-iannone]

Agreed, we could definitely make this faster/better. (Part of the entire scan_data() overhaul.)

[kmasiello]

Model improvements to scan_data() off of the python pointblank's DataScan to create a LLM-ingestible summary.