MLP TNN tuning (#107)

MLP TNN tuning (#107)

* Reimplement tnn integration in MCTS

* Correct timing bugs; tuning

* Makefile

* Do not quit when NN is not found

* Correct bugs in float parsing and NN priors

* New neural networks of varying focal dst

* Solve many TODOs

* Fix bug with parsing; data-set only use winner moves

* Limit knowledge collection to game plays by the winner

* Correct and improve game record details and add endpoint to display game info

* Pass instead of playing if overwhelming - in text mode

* Improve int parsing

* Improve parsing at parameter setting

* Minor changes to collection of knowledge

* Minor fixes; improve board guessing; more neural nets

* Add new neural networks; correct permissions

* Fix bug in finding opening book files

* Fix opening book writing

* Add options

* Version 1.24

* Correct script permissions

* Store even priors halved; improve first executions of MCTS on --benchmark runs

* Changes to benchmarking

* Comments

* Update scripts with new --losing flag

* Remove unused cpp definitions