Compact Table optimization

[PierreTsr]

This is the new constructor proposed in #137 with the associated testset.

I first changed the function buildResidues as some supports weren't present in the table when initializing the constraint. But then I figured it would be more efficient to give the constraint already cleaned versions of the supports and table.

The counterpart is that between different instances of the constraint sharing the same instances of table/supports, the variables should initially have the same domains (otherwise the support will be setup incorrectly for some instances).

To sum up, this new constructor gives much more freedom to the user but should be well understood in order not to tigger errors.

Closes #137

[qcappart]

Sounds good to me ! It will be interesting to access the memory reduction that you gain with this optimization