Vary time use with eval.

[xoto10]

Adjust time use depending on the current eval.

Passed STC :
LLR: 2.94 (-2.94,2.94) <0.00,2.00>
Total: 101696 W: 26651 L: 26238 D: 48807
Ptnml(0-2): 400, 11602, 26459, 11959, 428
https://tests.stockfishchess.org/tests/live_elo/660187a50ec64f0526c557f6

Passed LTC :
LLR: 2.94 (-2.94,2.94) <0.50,2.50>
Total: 60648 W: 15550 L: 15187 D: 29911
Ptnml(0-2): 40, 6356, 17171, 6715, 42
https://tests.stockfishchess.org/tests/live_elo/660298ed0ec64f0526c566d0

Values were found using two tunes with the final values taken from the ltc tune after 62k games :
stc - https://tests.stockfishchess.org/tests/view/65fb526b0ec64f0526c50694
ltc - https://tests.stockfishchess.org/tests/view/65fd36e60ec64f0526c5214b

Ideas for future work;

• tune these values with the other TM adjustments
• try narrower bands
• calculate adjustment for exact eval by interpolation

Bench 2103324

[xoto10]

Notes:

• I changed the variable name from the one used in the tests and made evalLevel[] a static constexpr
• I ran clang-format and pushed again

[Disservin]

evalLevel should be named in PascalCase :)

[xoto10]

evalLevel case updated

[mstembera]

It's surprising that the values in EvalLevel[] don't seem to have a trend predictable from bestValue.

[xoto10]

Yes. Maybe it could be approximated, we could try smaller divisions to get a better idea of the shape. Seems to be neutral when eval ~= 0, lower when slightly away from zero and then swinging to higher time use when the position is more critical (roughly eval > 120cp).
I expected it to be symmetrical around 0, but it looks like it isn't.

[xoto10]

I screwed up there, updating the branch with the values after the LTC tune finished, I've reverted it back so that this PR has the values which passed stc and ltc. Currently testing the 100k values using the timetotal3c branch.