Bug fix to hinge() and check of data added

[oharar]

I've made the following changes:

1. hinge() now doesn't return columns with the same features.
2. maxent() checks to see if any column in data only has a single unique value, and stops if it does (requested here: maxnet() sometimes fails with hinge features #26).
3. Minor update to documentation to stop R CMD check complaining

This code shows the changes:

library(disdat)
library(maxnet)
sp <- "nsw38"

bgEnv <- disBg("NSW")
EnvNames <- names(bgEnv)[!(names(bgEnv)%in%c("siteid", "spid", "x", "y", "occ", "group"))]
Pres <- disPo("NSW")[disPo("NSW")$spid==sp,c("occ", EnvNames)]

dat <- rbind(Pres, bgEnv[,c("occ", EnvNames)])
dat$vegsys <- factor(dat$vegsys)

This now works. Before it was causing a problem in predict.maxnet()

MaxNet.mod <- maxnet(p=dat$occ, data=dat[,EnvNames],
f=maxnet.formula(p=dat$occ, data=dat[,EnvNames],
classes="h"))

dat$One <- rep(1, nrow(dat))
EnvNames1 <- c(EnvNames, "One")

This now gives an informative error

MaxNet.mod <- maxnet(p=dat$occ, data=dat[,EnvNames1],
f=maxnet.formula(p=dat$occ, data=dat[,EnvNames1],
classes="h"))

[oharar]

Added option to change weights on absences: these can be a scalar or the same length as p. This means we can use MaxNet on presence/absence data, and also test how close 100 is to infinity (or even use area based weights)

[oharar]

Now if any element of the entropy has p=0, so p log p = -Inf, removes that element (=sets to 0) with a warning. Resolves #28