Make triangulations unique to fix refimages issues. Also tweak examples used for refimages

[DanielVandH]

Description

Passes randomise=false to all calls to triangulate to force them to be unique.

I have also tweaked the refimage test examples so that issues with randomisation or iteration order across Julia versions hopefully won't break them anymore. This hopefully fixes the issues you were having @SimonDanisch

Type of change

• Bug fix (non-breaking change which fixes an issue) (kinda)

Checklist

• Added an entry in CHANGELOG.md (for new features and breaking changes)

[DanielVandH]

Are the tests failing because the refimages aren't updated yet, or they actually just failing because I've done something wrong here?

[asinghvi17]

Looks like the reference images need an update because of this PR's changes, which makes sense.

[SimonDanisch]

Have you checked that all tests look correct? Then I could upload the delaunaytriangulation related reference images.

[DanielVandH]

When I made the PR at the time everything was correct. I can double check again, I just wasn't sure if any of the other changes since had updated the refimages as I'm not sure how that's done on your end.

I'll check again soon and let you know once I've confirmed.

[ffreyer]

refimgs are partially updated for #3958 atm so some things will fail on master

[DanielVandH]

I've double checked and everything seems fine. I added another randomise=false just for a bit more protection. There is a failing test

CairoMakie/Lines from outside.png    

but I don't imagine that's related.

[ffreyer]

Yea that's a failure due to the other pr. The pr fixes a problem with how CairoMakie handles color interpolation with clipped lines. FastPixel should also be failing because of the other pr.

[ffreyer]

Two ref images changed in point density:

before	after

This changed in terms of triangulation:

And here it seems the colormapping changed, but I guess this could also be a triangulation change?

The new images look consistent to me across backends and the ones I didn't mention should be the same as before. I think all of these are fine/desired changes. @DanielVandH If these still look good to you I'd be happy to merge this

[DanielVandH]

Those all look as intended to me! Thanks.