Rewrite pfold to fix sample counting issue & epfold to include more statistics

As implemented pfold did not count the contributions to each phase bin correctly.
Now a fractional estimate is done which should set the weights correctly.

epfold now includes lstat for events and light-curve. Also z^2 statistic was added for
both light-curve and events, events are based on the lc formulation. The original
z^2 version was done for events (and probably is a bit more sensitive for events),
I believe it not to be of strong impact.