refactored temperature schedule and added new methods

[nmamano]

• getPBad termines when thermal equilibrium is detected (this is almost instantaneous for high temperatures, thus speeding up linear regression by a factor of 2)
• unified all code based on doing linear search in logarithmic space (getAcceptableTFinal, temperatureBracket, ...) into a single method called "doublingMethod"
• added a binary search method based on pBad. This is the binary search method that makes sense to compare against linear regression, as opposed to the old statistical-test method that was doing something else.
• added the method from the paper "Computing the Initial Temperature of Simulated Annealing" by Ameur. (still a bit finicky)