Fix documentation for qmu_tilde test statistic

[kratsg]

Looking at the docs, it seems μ^>μ is the only time we can get q~μ=0 (whether asimov or not).

The docs don't actually match equation (16) in the paper, where the first line is $\hat{\mu} \leq \mu$.

Originally posted by @masonproffitt in #1992 (comment)

As it turns out, the docs don't match the code:

pyhf/src/pyhf/infer/test_statistics.py

Lines 30 to 32 in cbd68c8

	qmu_like_stat = tensorlib.where(
	muhatbhat[pdf.config.poi_index] > mu, tensorlib.astensor(0.0), tmu_like_stat
	)