ENH: hypothesis test that are robust to I(0) and I(1) residuals

[josef-pkt]

based on a PR for scipy.stats
scipy/scipy#10985

There are articles for trend and structural break testing that are robust to both short and strong correlation, i.e. I(0) and I(1) residuals.

Vogelsang, but there are additional more recent articles.

main problem is non-standard asymptotic distribution
The test statistic looks like a weighted average between distribution under I(0) assumption and the distribution under the I(1) assumption.

[josef-pkt]

We now have permission of Vogelsang to use, translate his code, @solomonvimal obtained it from Vogelsang

Location for test: I think it fits best with tsa hypothesis tests in tsa.stattools
unit root tests allow for trend
trend test allows for unit root

@bashtage