Random order for tests #7089
Description
Summary
Run tests in random order. By randomizing the classes and the the tests within them, as well as the non-parallelizable tests.
Allow reporting and consuming the random seed that is used for the run, to allow repeatable order of tests.
Consider the impact on dotnet test, where we probably want to define one seed for whole solution rather than random seed per project.
Background and Motivation
Running tests in random order can be useful tool to figure out hidden dependencies between tests. Also race conditions in code that are not normally found might be made visible by randomizing the tests.