0 value handling

[JuliaS92]

Major changes:

• added raplacement of ==0 values with nan as optional processing step (replacing step in data completeness filter)
• added dropping of none finite features as optional processing step (replacing dropping of all 0 step)
• rewrote the ttest to always operate on log transformed data for fold change and p-value calculation (p-values were calculated in untransformed space if not done durign preprocessing)

Tests were updated accordingly, but subset output changed due to the dropping of values. (Before, rows of nans and 0s were retained). Since this step is also used in batch correction the result changed.

TODO: Figure out which undisplayable p-values make it to the volcano plot function, as the axis of the fold change is wider than it needs to be.