NNUE evaluation threshold

[MJZ1977]

The idea is to use NNUE only on quite balanced material positions. This bring a big speedup on research since NNUE eval is slower than classical eval for most of the hardwares and specially on unbalanced positions with LazyEval.

STC: https://tests.stockfishchess.org/tests/view/5f2c2680b3ebe5cbfee85b61
LLR: 2.95 (-2.94,2.94) {-0.50,1.50}
Total: 3168 W: 560 L: 400 D: 2208
Ptnml(0-2): 21, 294, 819, 404, 46

LTC: https://tests.stockfishchess.org/tests/view/5f2c2ca6b3ebe5cbfee85b69
LLR: 2.98 (-2.94,2.94) {0.25,1.75}
Total: 3200 W: 287 L: 183 D: 2730
Ptnml(0-2): 4, 149, 1191, 251, 5

Bench 4746616

[vondele]

First ~10 Elo patch, quick!

[ssj100]

Just a query please - anyone can work out why SF suddenly becomes blind (and literally never finds the move) in "difficult" positions like the Gusev-Averbakh Queen sac position (sorta leads to zugzwang stuff)? SF NNUE prior to this patch found that move very quickly (often within seconds to a minute on 1-core), while regular SF never found the move. Since this patch got merged, SF NNUE goes back to never finding it. I also noticed some "old habits" of regular SF in other test positions, despite the evals being well below +1. I thought this patch only kicks in when evals are >2-3?

[vondele]

It is active for any value of the eval of the search, but only for positions with large(r) material imbalance.