Added scan for MPP in both directions, invert starting direction each time, dynamic dAngleMax

[ilario]

Here I join three rather big changes, if you prefer I can submit separated PR, depending on which of these changes you agree on.

1. the scan now looks for a MPP in a forward scan and for another in a reverse scan, uses the average of these two as the real MPP
2. the "walking to" steps are not used for MPP search, just the scans
3. the scan starts each time in a different direction, 'cause also this has an influence in the resulting MPP position
4. a dAngleMax of 25 degrees means that each scan is rather big, triggering more hysteretic effects and taking more time, here I implement a dynamic dAngleMax which decreases (the minimum is limited by dAngleMinLimit) as the Vmpp gets stable

[ilario]

This works very well with silicon reference cell (obviously)

(voltage is black, current is red, efficiency is green)

in perovskite cells just improves the measurement quality reducing the hysteresis triggered by a smaller scan

but further improvement is needed, see also #12