Clarifications and documentation improvements in iodata.basis

[tovrstra]

(This was a part of #146.)

There are some points of confusion in the documentation of iodata.basis and some names may be confusing. I'll maintain a list of suggestions for improvements below.

• Shell.nprim: The "number of primitives" can be interpreted in a few different ways. E.g. nexp for number of exponents has less ambiguity. contraction_length is too long. See Rename nprim to nexp where appropriate #357
• Shell.ncon: It should be clarified that different contractions within one generalized contraction may, but not must, have different angular momenta. They share the same exponents but have different contraction coefficients. See Clarify the representation of (generalized) contractions in docstrings #359
• MolecularBasis.get_segmented. The terminology of loop "unrolling" in the docstring is confusing. See Add prepare_segmented function #373

(Suggestions are welcome.)

Some of the points above imply minor API changes.