Add test for the following functions from csp.py

* revise
* AC3
* first_unassigned_variable
* num_legal_values
* mrv

* Update learning.py

* Added remove_examples function

Make load_MNIST function easier to read

In learning.py, I added the Hamming Distance metric.

* Rearranged Tests

- test_ngram_models to the top
- added test_viterbi-segmentation
- removed test_unigram_text_model

* "test_ngram_models" to "test_text_models"

* Improved BlindDog example

* Added 2D GUI IPython capability

* Demonstrated 2D Environment with GUI

* allowing import without ipythonblocks installed

* added tests for Direction

* fixed Direction arithmetic error

* Delete fig_5_2.svg

* add fig_5_2.png

* Update games.ipynb

* Update text.ipynb

* Update text.ipynb

* Update some comments

* Cleanup and remove some duplicate initialization

* Fix some comments quotation and little bugs: raising Exception instead of raising Failure, random.randrange(a,b) instead of random(a, b)

* Fix some comments quotation and remove explicit inheritance from object

* Fix 'r' in front of the ```parse_csv``` comments

* Fix quotation in comments

* Fix quotation in comments and fix bug in 'KB_AgentProgram'

* Update agents.py

* Update test_grid.py

* Update grid.py

* Update utils.py

in pseudo code the sequence of arguments is " WEIGHTED-SAMPLE-WITH-REPLACEMENT(N, S, W)"  same must follow in function "particle_filtering" in the file probability.py

* Update learning.py

weight_sample_with_replacement sequence of args

* Update probability.py

 weighted_sample_with_replacement sequence of args

* Update search.py

* Added Plurality Learner Plot Image

* Update learning.ipynb

Add tests to the following methods from CSP class

* result
* goal_test
* support_prunning
* suppose
* prune
* choices
* infer_assignement
* restore
* conflicted_vars

* Add flake8 config file

* Fix flake8 for test files

* Update Comments on learning.py + PluralityLearner

- Fixed some capitalization, spelling and quotation mistakes in comments

- In the PluralityLearner function, the nested function "predict" always returns the same output for a dataset, without taking into account the input ("example"). I defaulted the input as an empty list, so that we don't have to create (or find) a dummy example when we want to simply find the most popular class.

* Update learning.py

* Update learning.py

* Update learning.py

* Made Requested Changes

* changed mean boolean error

* Update learning.py

* fix bold look on github's ipython view

* standardise docstrings

Stop using statistics.mode, because it throws an error when there are ties.