Create stricter tests for jnd2xyz()

[Bisaloo]

The first commit of this PR highlights the fact that distances are not preserved in the 2D case. This happens because we rotate in the 3D space and project back in 2D.

The following commits adjust jndrot() to rotate in the 2D plane for the 2D case, and set default for axis1 and axis2 in accordance with this change.

TODO:

• update the docs of axis1 and axis2 to follow the changes made here

[Bisaloo]

It seems that there is a separate issue for dichromats, where distances returned by jnd2xyz() do not match the distance one would compute directly on coldist() output.

Since it has a separate cause, I will address it in a follow up PR and I have updated the tests to focus exclusively on the rotation aspect (via jndrot() for now.

[Bisaloo]

pkgdown workflow failure is related to r-lib/pkgdown#2802.

[thomased]

Thanks for this — great catch, and it looks spot-on to me. Nothing to suggest in terms of changes, happy to roll this in when you're ready.