Output of --bulk-debug-python generated by robust comparison script should include configuration values

[oleksandr-pavlyk]

nvbench-compare-robust supports --bulk-debug-python <PYTHON-FILE> option to generate Python script encoding paired benchmarks/axes.

It is very useful for sleuthing comparison status, making visualizations (see NVIDIA/cccl#10636 (comment)), etc.

It is useful to include information about decision tree configuration data in that file so that visualization could compare actual metric to threshold value.