Endgame tweaks for better explainability and ease of use

[domino14]

The endgame has some surprising behavior. For example, running an endgame at 4 plies does not always find the best 4-ply sequence. The following is an example of one such situation:

   A B C D E F G H I J K L M N O     ->               úrsula  LNOORVY  372
   ------------------------------                    arcadio           462
 1|=     '       =     A D D E R |
 2|  B R O W N       P R O E M   |   Bag + unseen: (5)
 3|    -   I   '   M I     L     |
 4|'     A N T I Q U E S - U N ' |   A A I S U
 5|        T           - A X e   |
 6|  "     L " G     J   N E T   |
 7|    ' B E G E T ' E   T ' S   |
 8|V I C E     C O L E   H   U P |
 9|    '       K   '     O ' R E |
10|  "       W O m A N E D   F A |   Turn 21:
11|H A F I Z E S       - I     G |   arcadio played E11 .ITI for 13 pts from a
12|Y     - I     '       A     ' |   rack of AIIISTU
13|D   -   T   '   '       -     |
14|R -     I "       "       -   |
15|O     '       =       '     = |
   ------------------------------

The best sequence here is: V(I)OL 3D is found. If opponent plays 15C AAS, then we play 10B YO and opp goes out with 13A (D)UI.

However, a 4-ply search will find VIOL/AIS/IRONY. You have to look 5 plies to find that AAS actually is better than AIS from the second player's perspective. See this Discord thread on it: https://discord.com/channels/741321677828522035/1157118170398724176/1281464776152842282

We need:

• Better documentation/explainability around this
• An option in LazySMP to not search different ply depths in different threads. This causes bugs (See endgame -multiple-vars not sorting properly #359) around sorting of multiple variants since the behavior is nondeterministic and it's finding better plays
• A higher default endgame ply count. (Default is 4)
• Sort multiple variants by value prior to presentation, in case the lazySMP different ply depth search is turned on.