Positive decomposition

[WoutPiessens]

Building on #924 and #929, this is a first implementation of the separate positive decomposition for global constraints.
For now:

• decompose_positive() is called if the global constraint c is top-level or positively reified (b => c). In case the global constraint is nested in any other way, it is decomposed as it was before.
• Only the MDD global constraint supports a separate positive decomposition

[tias]

reviewed in #981

[tias]

neat.

One change needed I think.

Are we sure the 'constraining','defining' is valid for the -> of MDD? That was an open question previously; is there a way to add a test for that?

[WoutPiessens]

neat.

One change needed I think.

Are we sure the 'constraining','defining' is valid for the -> of MDD? That was an open question previously; is there a way to add a test for that?

The constraining & defining constraints for positive decomp do seem right, based on my intuition and also on the tests

[tias]

OK great, it also makes sense to me but good to see it tested.

I rewrote the 'if' part slighly to have clearer non-overlapping cases