Enhancement

Current executor test framework passes expected columns to execute_streams() and does equality checking within it. This approach has some drawbacks

1. hard to recognize which line of the unit test file causes failure. Note that, current implementation will always output the file name and line number of a single internal function.
2. hard to do different kinds of equality checking. To be specific, with higher concurrency, the final columns result might not be stable (the order might change). However, current implementation lacks the ability to handle this situation. And current workaround (appending a topN executor when results might be unstable) is not elegant. (we should keep each unit test as self-contained and less dependent as possible)