Provide more extensive documentation

[zain-sohail]

We start with #474 to use a different theme. This allows very easy way to write docs in markdown (is also currently supported but doesn't format as nicely).

Todos:

• Move the dataset module guide from api to user guide
• Provide an IO user guide (For example, sed.io provides a very easy way to save and load binned data in hdf5, which can then be further used)
• Guide to calibration inner workings (can use Energy calibration bias shift #411 description for energy)
• Add more here

[rettigl]

Some text for energy calibration description:

Generally, the calibration estimates the functional form Energy(TOF), based on the TOF values for a given feature at different energies (bias voltages). This is being done either by nonlinear least-squares fitting, or an analytic solution of a polynomial approximation.
As the energy values provided are Bias voltages, and not directly related to the energy scale to be calibrated, this leaves a constant energy offset free to be determined. The old method allowed to define a specific "reference trace", i.e. a specific bias voltage, and the energy of the reference feature at that particular bias voltage. Ideally, this should have been the bias voltage of the actual measurement, or a corresponding additional shift needed to be applied.
The new approach is to shift the calibration by the bias voltage of the calibration (here, any one can be used, per default the first. Hence no need for ref_id anymore.), i.e. to calculate the energy scale at 0V bias voltage. When applying the energy axis, now the actual bias voltage of the measurement is being added, to obtain the correct energy scale.