difference in state of index variable for uncompiled and compiled for loops at end of loop

[paciorek]

Looks like C++ increments the index variable after the loop iterations, so we have a subtle, small difference in compiled/uncompiled behavior. This affected the IF2 continueRun method where our logic assumed R-like behavior (now fixed in nimbleSMC devel branch).

 nf <- nimbleFunction(
     run = function() {
         returnType(double())
         for(i in 1:5) {}
         return(i)
     })
cnf <- compileNimble(nf)
nf()
# [1] 5
cnf()
# [1] 6