Python MNSIT example: use better topology

SP had dimensions of (X, 1)
Changed to (sqrt(X), sqrt(X))
This helps because the input images are 2-D

Measurements:

 Num Evals | Mean Score | Parameters
        44 |   0.9542   | Before
        20 |   0.9554   | After

P-Value of null hypothesis: 0.000793

py/htm/examples/mnist.py

@@ -81,7 +81,7 @@ def load_images(file_name):
 default_parameters = {
     'potentialRadius': 7,
     'boostStrength': 7.0,
-    'columnDimensions': (28*28*8, 1),
+    'columnDimensions': (79, 79),
     'dutyCyclePeriod': 1402,
     'localAreaDensity': 0.1,
     'minPctOverlapDutyCycle': 0.2,