Make `grid` a product of two grids

[albop]

This is a first step towards #188 .
The idea is simply to make model.grid a cartesian product of two grids. A product grid could attach names to each factor as in grid = GridProduct( {'endo': endo_grid, 'exo': exo_grid} ), then one could be able to access grid.endoandgrid.exoinstead of the currentmodel.endo_gridandmodel.exo_grid. This would clarify that a decision rule is defined on one single grid. An unresolved issue is how to pass discretization information to solution routines. I have been thinking that a natural to do so would be to do model.domain.discretize(n=[10,10])instead of the currentmodel.get_endo_grid(n=[10,10])`.

[albop]

Done.