revisit ordering of loops in declaration functions

[paciorek]

This shouldn't matter except for very large matrices/arrays, but if we have a matrix stored column major, we would want the looping in the declaration functions to loop over the rows in the inner loop.

If one writes a model like this:

for(i in 1:n)
  for(j in 1:m)
    y[i,j] ~ dfoo()

then the looping at the moment will be over columns in the inner loop.

This is presumably a minor efficiency issue, and we should not focus on it now.

[perrydv]

That makes sense. There are several small ways we could make the iterations in C++ more efficient (why I like having them hand-coded instead of purely generated). One is to iterate through elements of y rather than indexing into it every time. That could make more sense if we know we have the loops ordered for sequential access of y's memory. But I agree we want to make things work for now and revisit for efficiency later.