BUG: ma.median alternate fix for #7592 - #7635

@ahaldane Ping.

Either solution (this or #7592) seems fine to me, at least with the suggested changes in either PR. On the plus side this PR avoids the convoluted isinstance checks, on the downside it introduces a separate branch to the code.

Let's not be like "Burdian's Ass" who can't choose a pile of hay!
This PR seems slighly less tricky, how about going with this one?

I think to tidy it up you should follow @charris' sugestions. I also think it might be better not to have a separate _median1D method, and instead simply move its code to be inside the if-block in _median, since with @charris's suggestion it is only 6 lines long.

I'd also suggest you change low = asorted[ind] to low = asorted[tuple(ind)] and same for high. Even though it won't change any of the results it reduces the chance of miscellaneous bugs down the line

(see also my comment above in @charris's thread)

Looks good to me! I would like to merge.

@AmitAronovitch one or two final things: I would get rid of the parenthesis in if (asorted.ndim == 1) (it's more pythonic without them), and maybe also change mean(0) to just mean() if that's correct.

Further, I plan to squash everything together if you don't do it for me. If you do end up squashing, could you format the commit message according to the guidelines? In the commit message you should also include the sentence "Fixes #5969." so that git auto-closes it for us.

I kept the casting to tuple out of the squashed commit, since it is independent of the bugfix.
Hope that's OK

Looks good, thanks a lot for all the revisions @AmitAronovitch!

Merging now.