New net trained using wdl 0.5 (#84)

Passed STC:
Elo   | 10.39 +- 7.05 (95%)
SPRT  | 8.0+0.08s Threads=1 Hash=8MB
LLR   | 2.90 (-2.25, 2.89) [0.00, 5.00]
Games | N: 5016 W: 1426 L: 1276 D: 2314
Penta | [90, 545, 1105, 661, 107]
http://aytchell.eu.pythonanywhere.com/test/413/

Something to note is that I needed to reconvert my dataset as the score was stm relative, Im pretty sure what I converted was the exact dataset with fixed scores, however it seems that a network trained with wdl 1 on the fixed data set loses a few elo: http://aytchell.eu.pythonanywhere.com/test/408/

bench 5435503

src/nnue.cpp

@@ -18,7 +18,7 @@
 #undef PUSHED_MACRO
 #endif
 
-#define defaultNetPath "src/Nets/moly_20240501.nnue"
+#define defaultNetPath "src/Nets/moly_20240526.nnue"
 
 INCBIN(network, defaultNetPath);
 const Weights defaultWeights = *reinterpret_cast<const Weights*>(gnetworkData);