⚖️ Re-train with a mix of WDL and eval

[eduherminio]

Attempted with wdl-score-mix-base branch of https://github.com/lynx-chess/texel-tuner, see lynx-chess/texel-tuner#11 commit 3a301972e5dd153b659896ed9f4710338820d192.

No WDL

Test  | eval/wdl-score-mix-wdl-0
Elo   | 0.85 +- 2.27 (95%)
SPRT  | 8.0+0.08s Threads=1 Hash=32MB
LLR   | -0.33 (-2.25, 2.89) [0.00, 3.00]
Games | 44650: +13166 -13057 =18427
Penta | [1491, 5157, 8792, 5522, 1363]
https://openbench.lynx-chess.com/test/1382/

25% wdl

Test  | eval/wdl-score-mix-wdl-25
Elo   | 0.83 +- 1.89 (95%)
SPRT  | 8.0+0.08s Threads=1 Hash=32MB
LLR   | -0.49 (-2.25, 2.89) [0.00, 3.00]
Games | 63890: +18828 -18675 =26387
Penta | [2076, 7405, 12689, 7840, 1935]
https://openbench.lynx-chess.com/test/1384/

50% WDL

Test  | eval/wdl-score-mix-wdl-50
Elo   | -1.03 +- 2.74 (95%)
SPRT  | 8.0+0.08s Threads=1 Hash=32MB
LLR   | -2.26 (-2.25, 2.89) [0.00, 3.00]
Games | 30776: +9001 -9092 =12683
Penta | [1065, 3567, 6086, 3734, 936]
https://openbench.lynx-chess.com/test/1383/

75% WDL

Test  | eval/wdl-score-mix-wdl-75
Elo   | 0.77 +- 1.69 (95%)
SPRT  | 8.0+0.08s Threads=1 Hash=32MB
LLR   | -0.81 (-2.25, 2.89) [0.00, 3.00]
Games | 80302: +23695 -23517 =33090
Penta | [2633, 9395, 15728, 9951, 2444]
https://openbench.lynx-chess.com/test/1385/

90% WDL

Score of Lynx-eval-wdl-score-mix-wdl-90-5615-win-x64 vs Lynx 5612 - main: 5691 - 5644 - 8055  [0.501] 19390
...      Lynx-eval-wdl-score-mix-wdl-90-5615-win-x64 playing White: 4238 - 1523 - 3934  [0.640] 9695
...      Lynx-eval-wdl-score-mix-wdl-90-5615-win-x64 playing Black: 1453 - 4121 - 4121  [0.362] 9695
...      White vs Black: 8359 - 2976 - 8055  [0.639] 19390
Elo difference: 0.8 +/- 3.7, LOS: 67.1 %, DrawRatio: 41.5 %
SPRT: llr -0.542 (-18.8%), lbound -2.25, ubound 2.89