improvements to Dataset.select()

[Thomas Gilgenast]

we are interested in making Dataset.select() more powerful

• it should allow selection as a DataFrame, or as a numpy array
• it should allow selection including or excluding nan's

• it should allow selection of multiple columns at once, e.g. the call

  obs, dist = d.select(['obs', 'distance'], rep='v65_rep1', region='Sox2')
  

  should do something like

  obs, dist = d.df.loc[d.df['region'] == 'Sox2', [('obs', 'v65_rep1'), ('distance', '')]].as_matrix().T
  

[Thomas Gilgenast]

this is unlikely to get a fix because lib5c.structures.dataset.Dataset is likely to be superceded by a new data structure based on the hic3defdr data layout

this issue is still helpful for the discussion of the features we would like to see in that new data structure