[Doc] improve documentation for p-values

[JosiahParry]

localG_perm() lists three different p-values and it's not entirely clear which are which. My understanding is the following:

• Pr(z != E(Gi)): "standard" p-value under the assumption of normality
• Pr(z != E(Gi)) Sim: "standard" p-value using bootstrap simulation
• Pr(folded) Sim: conditional permutation that counts extremes on either tail

Can you confirm how each p-value is calculated? I'm happy to make a PR based on informal clarification here. Additionally, it may be worth having a single Rd file for p-values that these various LISA stats can point to.

[rsbivand]

Hi @JosiahParry : this is associated with https://github.com/pysal/esda/issues/199 which you started. We'd like to stay aligned, and probably in addition move towards using Luc's "interesting" rather than "significant", as per https://r-spatial.org/book/15-Measures.html#local-morans-i_i - the chapter is the documentation (more or less). Has esda gone "live" on this? I'm unsure about rgeoda and the GeoDa family on this.

All the outcomes you list are two-tailed and based on conditional permutation, the first is (Gi - mean(sims))/sqrt(var(sims)) which works well when the data are bell-shaped, the second is ranked, and the third is ranked but mimics the folding of the two tails into [0, 0.5] as in esda and GeoDa.