False positives wherever we index with settings.

We need more arguments in parameterized scientific functions,
more methods in Qt subclasses.

Functions map and filter can be nice and even clearer than
comprehension in some cases, e.g.

    max(map(len, rows), default=0)
    filter(len, transactions)

vs.

    max((len(row) for row in rows), default=0)
    (t for t in transactions if len(t))

Numpy has a unique dynamic way of loading its members which causes false
positives for pylint. Best option is to disable this check, since any
bad imports will fail the unittests anyway.

This function turns the pylint score into a normal issue count, which
can be used for comparing code style on pull requests. Relying only on
code score is not good, since it is possible to introduce new issues as
long as there are enough new lines of code without issues.

This reverts commit f5963a24.