Add base dcm class

[Illviljan]

Example attempt to align return types for values.

The API is inspired by numpy and xarray.

High level ideas:

• Inherit/use only a single dcm class, reduces typo risks etc, Risk of getting too much clutter.
• Follow typical conventions popular in data analysis tools like numpy /xarray
  • Use self.shape instead of x_dimension y_dimension. -> Following numpy.array.shape standard.
  • Use list of list for WERT values. -> Easy to initialize numpy.array then.
  • Use tuple of lists for axis values. -> Matching order of numpy.array.shape-tuples
  • Move remaining metadata to a self.attrs dict. -> In similar fashion as xarray/pandas
• A lot of type hints to clarify intention of methods and attributes.

Only maps works in this form for now. I want some discussion before going all in on this.

References:
https://www.etas.com/download-center-files/products_ASCET_Software_Products/TechNote_DCM_File_Formats.pdf

[dkuschmierz]

Wouldn't it be better to directly use numpy within the project instead of generating compatible interfaces?

[Illviljan]

That's ok as well. I was just thinking dcmReader didn't wan't to have numpy as a dependency. Should numpy be a dependency?