Fine performance metrics: documentation etc.

- Part of #7665

Final users should be educated on
- how to properly read and interpret these metrics
- how to feed custom metrics into the system; e.g. to mark GPU time or custom I/O functions

Write user documentation and blog posts, record demos, etc.

I suggest to leave this task until after we have delivered enough of the above items that we feel the system is good enough for general use.

This ticket needs to be refined at a later date.