Improve the max turn Algorithm

[k0vl]

At F17, this project uses a naïve estimation of maxTurn, which is modified from http://unixpapa.com/floodit/, that does a linear scaling using a baseline of 25 turn with 14x14 board and 6 color.
maxclick= Math.floor(25*((nrows+ncols)*ncolors)/((14+14)*6));
This implementation generates unplayable board at small size, and overly easy board at large size.

We have since reworked the algorithm to perform random simulation before each game, and taking the average of the top simulations as the max turn. However it is still easy at large size.

Consider implementing a greedy algorithm that maximizes cells flooded each turn, and using it as the baseline.
point estimation: 200

Consider implementing a greedy algorithm that looks ahead several turns, which would overcome certain edge cases*.
point estimation: 400

Consider implementing a neural network.
point estimation: 700

*reference: https://kunigami.blog/2012/09/16/flood-it-an-exact-approach/