As we are considering replacing mergesort with another algorithm in hope of improving performance for nearly sorted data #12186 , maybe we should include benchmark for mergesort in the benchmark suite, which currently only takes care of quicksort.
As I see it, there are two ways to do this:

• Simply add 2 benchmark functions on random data and nearly sorted data for mergesort.
• Separate sorting algorithm benchmark from bench_function_base.py to a new file and construct a systematic benchmark framework for sorting algorithms. This can provide a useful tool for future optimizations, however it may introduce lots of unnecessary benchmark results.

Any advice or suggestions? Thank you all in advance.