[Update] ``Cartesian`` Module

Review cartesian module and update as required:

• line_profile to see if further numbafication is useful or not

Current discussion from @albop:
@oyamad , @mmcky : On the cartesian function, the reasoning was that the loops and list comprehensions are made over the number of dimensions, so that they are probably only a few iterations. Appart from the loop over _repeat_1d, it seems likely that everything else is going to stay in python mode, with little performance gain if any. This was just my feeling though, I didn't test it.

• update so that numba isn't an explicit requirement (numpy version?)
• Incorporate Issue #98

Start by reading the cartesian module, especially line_profile, and review the discussion about loop performance and numba. Check Issue #98 and identify the current numba requirement. Done means resolving the checklist items: deciding whether numba helps, removing its explicit requirement if appropriate, and incorporating Issue #98.