Fix compilation error with min/max under certain circumstances

[paciorek]

In the following case, we have an untrapped error (see issue #1533):

code <- nimbleCode({
    x <- max(0,y-z)
})
m <- nimbleModel(code,inits=list(z=-1,y=5), buildDerivs=TRUE)
cm <- compileNimble(m)

It occurs with if(iName == 'log' && (is.numeric(iArg) | is.logical(iArg))) in cppOutputNimDerivsPrependType. When the condition after && is TRUE, this errors if iName is NULL. If condition after && is FALSE (e.g., if the model code is x <- max(w,y), then execution continues.

I've wrapped the check in isTRUE as a rudimentary fix.

(Side note: max(a,b) is converted to pairmax behind the scenes, and I think this is why cppOutputNimDerivsPrependType, but I haven't fully grasped the details.)

[paciorek]

Also, just noticing we need to convert | to || to be consistent with R's handling of scalar versus vector boolean arithmetic (and I think R is getting more strict about this in some recent or future version).