Renamed deviance to log_loss in GBDT hyperparams

Scikit-Learn has renamed the `deviance` hyperparameter in their GBDT
Classifier to `log_loss`. This deprecation prevented GOSDT from being
built for Python3.10+. The `log_loss` change does force us to no-longer
support Python3.7.

Signed-off-by: Ilias Karimalis <iliaskarimalis@outlook.com>

gosdt/model/threshold_guess.py

@@ -4,7 +4,7 @@
 import time
 import random
 import sys
-import os  
+import os
 from queue import Queue
 import pathlib
 
@@ -16,7 +16,7 @@
 
 # fit the tree using gradient boosted classifier
 def fit_boosted_tree(X, y, n_est=10, lr=0.1, d=1):
-    clf = GradientBoostingClassifier(loss='deviance', learning_rate=lr, n_estimators=n_est, max_depth=d,
+    clf = GradientBoostingClassifier(loss='log_loss', learning_rate=lr, n_estimators=n_est, max_depth=d,
                                     random_state=42)
     clf.fit(X, y)
     out = clf.score(X, y)