Adjust LMR with correction history

[xu-shawn]

A positive constant increase in base reduction is applied to counter the decrease in average reduction from this tweak.

Passed STC:
LLR: 2.98 (-2.94,2.94) <0.00,2.00>
Total: 109216 W: 28415 L: 27989 D: 52812
Ptnml(0-2): 310, 12848, 27911, 13184, 355
https://tests.stockfishchess.org/tests/view/6760bb0e86d5ee47d9542f26

Passed LTC:
LLR: 2.95 (-2.94,2.94) <0.50,2.50>
Total: 66918 W: 17073 L: 16694 D: 33151
Ptnml(0-2): 33, 7175, 18666, 7550, 35
https://tests.stockfishchess.org/tests/view/6761e10f86d5ee47d95431fa

bench 1079115

[github-actions]

clang-format 18 needs to be run on this PR.
If you do not have clang-format installed, the maintainer will run it when merging.
For the exact version please see https://packages.ubuntu.com/noble/clang-format-18.

(execution 12405451763 / attempt 1)

[mstembera]

Congrats! Perhaps I'm a bit old school but I really dislike returning auto. Not being able to immediately know what is being returned unnecessarily eats up brain cycles while reading code. Curious what others think.

[xu-shawn]

Maybe it would be better to give correction score a type alias? that could also save the template for eval correction

[Disservin]

correction value is just a dumb int, no need for any template or big abstraction imo

[ehsanrashid]

why use std::abs()
r -= std::min(std::abs(correctionValue) / 32768, 2048);

is it not better to use the raw value ?

[cj5716]

std::abs tells us the magnitude of the correction (ie, how unstable the position is)