More specific Python error mapping

[flying-sheep]

I’d like to improve thrown errors. Ideally our tests would make sure that

• specifying some invalid index in __getitem__ always raises IndexError
• some invalid type passed to a parameter raises TypeError
• some invalid value of the correct type being passed to a parameter raises ValueError
• only resource errors or other things users have no direct control over raise RuntimeErrors

For this, I think .map_py_err should be the more common API that we use on known error types that we can cleanly map to Python error types, and .map_py_err_from_str is the (intentionally more verbose) variant inline mapping of

1. rare error types
2. instances of known errors that get raises in a context where they indicate something else happened, e.g. an error that would normally be an IndexError occurs due to some runtime issue and should become a RuntimeError in that case (like a catch-and-reraise)