Determining phase of a variable in decision

[rajdeep87]

Strategy to dynamically change the variable phase inside the decision heuristics.
This is relevant in the context of tpolyhedra heuristic ENABLE_HEURISTIC which biases the variable value to one extreme instead of a binary search. Such bias may be disadvantageous for some program. For example, learning0 regression suffers because of this heuristic.

[rajdeep87]

If the trail already contains a decision, say x<=10, then changing the phase of the variable x to generate a decision x>10 would lead to inconsistent trail. Hence, changing phase of a variable first needs to make sure that the decision level corresponding to x<=10 is already backtracked and the trail does not have contradicting elements.