⌛ Make soft and hard bounds explicit, simplifying the existing logic

[eduherminio]

Reimplementation of #447.
This is an attempt to standarize time management using the hard/soft bound technique, where different values are used to stop the search after each IDDFS iteration (soft bound) and to stop the search at any moment (hard bound).

We were already using this technique via applying DecisionTimePercentageToStopSearching coefficient to the hard bound to get the soft one.
Last tm revision was done in #288

I run this non-ref tests, but I'm trully merging this together with #665, I just want to make clear the separation between the implementation and the value tweaking (fixing even, lol)

8+0.08
Score of Lynx-time-management-soft-hard-bounds-2-2633-win-x64 vs Lynx 2630 - main: 1080 - 981 - 1318  [0.515] 3379
...      Lynx-time-management-soft-hard-bounds-2-2633-win-x64 playing White: 762 - 291 - 637  [0.639] 1690
...      Lynx-time-management-soft-hard-bounds-2-2633-win-x64 playing Black: 318 - 690 - 681  [0.390] 1689
...      White vs Black: 1452 - 609 - 1318  [0.625] 3379
Elo difference: 10.2 +/- 9.1, LOS: 98.5 %, DrawRatio: 39.0 %
SPRT: llr 2.91 (100.7%), lbound -2.25, ubound 2.89 - H1 was accepted

40+0.4, cancelled
Score of Lynx-time-management-soft-hard-bounds-2-2633-win-x64 vs Lynx 2630 - main: 401 - 399 - 640  [0.501] 1440
...      Lynx-time-management-soft-hard-bounds-2-2633-win-x64 playing White: 306 - 90 - 323  [0.650] 719
...      Lynx-time-management-soft-hard-bounds-2-2633-win-x64 playing Black: 95 - 309 - 317  [0.352] 721
...      White vs Black: 615 - 185 - 640  [0.649] 1440
Elo difference: 0.5 +/- 13.4, LOS: 52.8 %, DrawRatio: 44.4 %
SPRT: llr 0.32 (11.1%), lbound -2.25, ubound 2.89