pylint improvements / refactorings

[The-Compiler]

Currently working on some major pylint changes because we pinned it due to a bug (pylint-dev/astroid#772) but never unpinned it...

Some notes on stuff we should probably do at some point, but not right now:

• Use UPPER_CASE for enum members, like also suggested by the Python enum docs (this has confused contributors before IIRC)
• Use -style instead of -rgx in the pylint config: https://pylint.pycqa.org/en/latest/user_guide/options.html#predefined-naming-styles
• Consider using lazy formatting for logging output (on the other hand, we have debug logging to RAM by default, so with default settings, it won't change much)
• Decide what we want to do for "W0707: Consider explicitly re-raising using the 'from' keyword (raise-missing-from)", ignored for now (also see Add B904 - Use raise ... from err in except clauses PyCQA/flake8-bugbear#181 for some related discussion)
• Consider re-enabling some flake8/pylint checks which overlap - sometimes people only use one tool, and e.g. flake8 is easier to integrate. Doesn't hurt if both point the issue out.
• Review disable=consider-using-with ignores and see if we can restructure code to use with there too
• See if we can change test imports so we don't need to ignore wrong-import-order for tests (there is no known-local, see e.g. [WISHLIST] wrong-import-order: Add parameters to known_*_library (future, standard, third...) pylint-dev/pylint#876 and [REF] import-checkers: Use of isort pylint-dev/pylint#879)
• Reconsider enabling too-many-try-statements
• Revert workaround for unused-before-assignment once Regression used-before-assignment with type annotation pylint-dev/pylint#5342 is fixed
• Reconsider whether we want to disable use-implicit-booleaness-not-comparison for tests
• Maybe report the new false-positives to pylint
• Check if we want to enable consider-using-tuple or use-set-for-membership
• Maybe enable pylint.extensions.mccabe with [DESIGN] max-complexity=12, seems to catch some stuff pylint does not
• Maybe enable pylint.extensions.redefined_variable_type (also as preparation for more mypy coverage)
• Consider enabling more doc checks (mainly missing-param-doc, maybe missing-raises-doc, missing-type-doc, useless-param-doc)
• Consider using disable-next comments more
• Report unsubscriptable-object false positive for config.cache with -j16 to pylint