Cost/reward notation

[ibarrien]

In the literature, cost and reward are complementary (not equal!) and only one is usually optimized. In the python interface, cost is being minimized as the default optimization.
It might be good to replace:
"Cost/Reward (r): the reward received for action a."
With:
"cost: the cost received for taking action a"
because in the subsequent example we have:
action:cost:probability | features