feat: new class structure for a values

[gottschali]

This PR reproduces the structure of the b-value estimators suggested by @schmidni for a-values.

• Structure: analysis/avalue with estimators and test
• AValueEstimator as abstract class
  • ClassicAValueEstimator
  • APositiveAValueEstimator
• add estimate_a function to __init__.py
• AMorePositiveAValueEstimator
• Better tests (it would be good if someone with more domain knowledge did this)
• sanity checks. Are there different ones for a values? Probably include the binning test.

Added a common method _reference_scaling since both estimators shared code for scaling by m_ref and scaling_factor.

calculate is overridden for APositive to get the correct docstring.

To avoid too many merge conflicts, we kept the original implementation for now.
But later if should be removed.

[schmidni]

@gottschali thanks for starting. I'm taking over from here as I have to do a few more changes too :)