Implement policy learning task

[bruno-f-cruz]

Requires:

• Choice on each trial is given by the animal stabilizing its force within a range (min<force<max) for N seconds. In practice this can be done by running a min-max rolling window of N samples and thresholding the resulting value. This can be further simplified by operating on top of the paired boolean threshold output.
• Once a choice is made, implements feedback modes:
• Gradient (feedback is continuous based on Force_choice - Force_target)
• UnsignedGradient (feedback is continuous based on abs(Force_choice - Force_target))
• StepGradient (feedback is discrete (i.e. higher, lower))
• ReinforcementLearning (feedback is discrete and unsigned)

[bruno-f-cruz]

After implementing the accumulation mode, we should probably follow the "factory"-liek pattern used for the continuous feedback but formalize it to the outcome feedback case too.