Refactoring: Put data in separate attribute of RTDC_DataSet

[paulmueller]

To separate data from functions, the column data loaded from tdms files should be stored in a separate attribute of RTDC_DataSet.

• Introduce the RTDC_DataSet.events attribute that can be addressed like a dictionary (e.g. ds._events["time"] returns a numpy array.
• For the transition: Override the getattr method of RTDC_DataSet and pray that this does not break anything
• Design the new submodule such that other data columns can be added/computed dynamically (e.g. average grayscale value, image, contour, trace)
• Make this compatible with the proposed .rtdc file format (dictionary access of events columns of an hdf5) (New standard file format for RT-DC data sets using hdf5 ZELLMECHANIK-DRESDEN/ShapeOut#151)

[paulmueller]

To allow to compute new data columns (area_um, time, volume, elastic modulus, grayscale) dynamically

• Implement a class that checks prerequisites of a column and that allows computing it
• Check against that class in RTDCBase.__contains__
• Use that class in RTDCBase.__getitem__
• Cache the computed data?
• Move definitions from dclab.definitions.tfd there.
• Move computations from ShapeOut there.