GForce could be even smarter

[ben-schwen]

reading this inspires some stretch goals for future work:

• GForce for simple regression coefficients ((mean(x*y) - mean(x)*mean(y))/(mean(x^2) - mean(x)^2))
• sqrt(min(a)) could do GForce aggregation
• list(mean(x)/sd(x), mean(x), sd(x)) could only compute mean(), sd() once
• max(x) - min(x) could compute range() with GForce and then derive min, max (i.e., single pass instead of two)

Originally posted by @MichaelChirico in #7401 (comment)

#7401 is a nice start, but we could start caching and rewriting things more to spare some passes over data

[ben-schwen]

This issue is basically asking for 3 different things:

1. Outer transformations applied after to GForce functions as e.g. sqrt(min(a)) and mean(x)^2
2. Inner transformations applied prior to GForce functions as e.g. mean(x*y) and mean(x^2)
3. Caching better query optimization, like using single pass range(x) instead of max(x)-min(x) and caching results for list(mean(x)/sd(x), mean(x), sd(x))

Note that 1) and 2) could actually be with chaining either moving the outer transformation to the next chain or moving the inner transformation to the prior chain.