Idea: PV reductions

[Aleks Peshkov]

While hunting bugs in the code of reusing previous iteration PV longer than current iteration depth, I come with thoughts:
1) Part of previous longer PV is ignored, so it was redundant work from the beginning.
2) PV nominal depth longer then its base iteration (or recursive internal iteration) depth is misleading measure of quality for future reuse.
3) Well-known check, recapture and other tactical extensions are sound idea for shaping search tree in general but tend to lead to search explosion.
4) Instead of tactical extensions of the current PV itself we can reduce search depth of its all refutation variations subtree.
5) We can save the quantity of reductions of the current (partial) PV in the transposition table at the additional field.
6) We can use this new "reduction" field not only in search but also in TT replacement scheme. If otherwise equal nominal depth TT nodes have different "reduction" value, the less reduction the more valuable (represent more search work) the node is.