grid_exo and complete polynomials

[sglyon]

In #112 we extended the CompletePolyDR to work with any type of non-empty endogenous grid.

We should be able to do the same for exogenous grids, but I want to have a quick discussion as to how we should combine the endogenous and exogenous dimensions.

Let's take Cartesian x Cartesian as an example. The Cubic spline code would form the Cartesian product of each of these grids. Should we do the same for the complete polynomial case?

What about Cartesian x Random? WE could also do a Cartesian product between these two grids, but it isn't as obvious that we should do that. Same for things like Cartesian x Smolyak

[albop]

@sglyon : this is a good question but I don't think the discussion will be quick...
Indeed nothing say that the right product to do is cartesian. We've already discussed the possibility of doing a Smolyak product of two Smolyak grids.
If random points are correlated to exogenous process, it might indeed pose some problems.
Also we might think that the extend of state-space could depend on exognous process.

Many reasons to keep this issue open...