[FEATURE] Implement a Texel Tuner for Evalulation

[DeveloperPaul123]

It would be good to make a tuner so we can auto-tune our evaluation values, specifically the PSQTs.

It seems that the gist is we create a gradient decent algorithm that takes in evaluation terms as a large vector of parameters. Then we minimize the predicted probability of the game result. We will start with using filtered data so that we can use the static eval function directly instead of having to use qsearch.

Data sets:

• gedas_filtered_sf_d9
• Lichess Big 3 (filtered)
• Zurichess

References:

• https://github.com/maksimKorzh/wukongJS/blob/main/docs/TEXEL'S_TUNING.MD
• https://github.com/maksimKorzh/wukongJS/blob/main/tools/eval_tuner/eval_tuner.py
• https://www.chessprogramming.org/Texel's_Tuning_Method
• https://youtu.be/3JCWxH6IehQ?si=zRNel4ggXx8jj0SE
• https://github.com/jw1912/hce-tuner/