There are several pieces of information that some files contain, others do not, examples are:

• # -*- coding: utf-8 -*- at the beginning of a file
• License information, eg # License: BSD 3 clause
• List of (some main) authors for that file, # Authors:

For the following reasons, I'd like to remove the encoding info, and have a consistent policy for the other 2 points:

• As a (newish) contributor, it's hard to figure out what is expected.
• Having those infos consistent in all files might help a tiny little bit to generate trust in the production-readiness of scikit-learn.

I haven't figured out a convincing policy for authors, maybe leave it as is.

Note that changes of those lines of code shouldn't produce merge conflicts, cross referencing #11336.