Include interval ends for sharpless, roundness, and peakmax

[mcara]

Closes #1977

This PR modifies selection of sources by shapness, roundness, and peakmax to include the interval ends, i.e., that sources that have roundness 0.0 are selected instead of being rejected when sharplo=0.0.

[larrybradley]

@mcara Do you recall what the old IRAF routines use? This is a breaking change, but I suppose could be classified as a bug.

[mcara]

Sorry, I replied in the issue: #1977 (comment)

[mcara]

This is a breaking change...

This could break some idealized tests, just like the one I am using in #1977 but in practice, what is the chance to have a perfectly round star (or a star with ellipticity precisely 2) in real observations? So, I believe this PR will have a minor effect, except for helping avoid this kind of tricks: https://github.com/spacetelescope/jwst/blob/6d0c2d50c72d32f9a16bf30264d3528bff53e36e/jwst/tweakreg/tests/test_tweakreg.py#L99 which seem unnatural given that the lower limit for roundness is 0 (it cannot be negative in the original IRAF code) yet I have to set this limit to a negative value in order to include 0.

[larrybradley]

LGTM. The old IRAF tools also allowed values at the limits.

Could you also please add a test for DAOStarFinder?

[larrybradley]

This also needs to be rebased due to a conflict in the changelog.

[larrybradley]

Thanks, @mcara. The test failure is unrelated to this PR.