Fix custom benchmark plotting for selected portfolio columns

[Whning0513]

Fixes #726.

plot_cum_returns() handled a custom benchmark_rets input by broadcasting it against self.obj, but Portfolio doesn't have that attribute. That raises an AttributeError as soon as you pass custom benchmark returns while plotting a selected column.

I changed the broadcast target to self.wrapper.dummy(group_by=group_by), which matches the portfolio shape the method is about to select from. I also added a regression test that exercises pf_multi.plot_cum_returns(column="a", benchmark_rets=pf_multi["b"].returns()).

Verified with:

• python -m pytest tests/test_plotting.py -k "cum_returns"
• direct repro script from the issue now returns successfully instead of raising AttributeError

[polakowo]

Thanks, the fix works. To make your test consistent with the surrounding tests, please shorten the name of the test method or include the test in already existing test methods, and I'll merge the PR.

[Whning0513]

Updated the regression to live alongside the existing cum-returns plotting tests and shortened the test name to match the surrounding style.