signal: Clarify behavior of *ord functions and improve test documentation

[endolith]

As described in #8397 (comment):

How did you derive such precise values for the expected results? Add a comment [in the test code] explaining where these numbers came from, so future maintainers can reproduce those expected values.

Or the values should be set to only test whether it meets the actual spec?

But in many cases, there's "wiggle room" in which the filter can vary while still meeting the spec, described here: #3219 (comment)

(buttord's hugging behavior was changed in #3235 (comment) )

The behavior of whether it will hug the inner or outer constraint should be defined in the docstrings, and it would be good to test this to make sure it stays consistent.