Add tests to ensure Cyberbrain has consistent behaviors across different Python versions

[laike9m]

Since we now put expected values in golden files, it's hard to spot different behaviors between different Python versions, which, is often an indication for potential bugs.

In theory, behaviors that are visible to people (what events a program has, their values, lineno, etc, but not including offset) should be consistent. And we need a test to make us feel confident about it.