Regression tests for tag order on the same object

[HolgerJeromin]

Came up in #3434 (comment)

it would be desirable to have an mechanism where we can formally define which of two conflicting tags wins the rendering. Otherwise these issues become forgotten after a while, and resurface as a side-effect once some layering etc is modified for other reasons.

This would be a classic case for an automated regression test. So if a refactoring changes the order travis for example breaks. If this is a desired change the author can adjust the test to the new order.

First we need a framework to make this automated check happen.
After that a list of desired tag order (shop vs amenity) could be build up.

[kocio-pl]

Do you have the idea what could be this framework or for some other details?

[jeisenbe]

This is related to #3880 - Symbol and label prioritization