Improve ``imod.mf6.NodePropertyFlow.from_imod5_data``

[JoerivanEngelen]

As @Huite mentioned, the current approach taken in computing k, k22, and angle1 from iMOD5's ANI package input is suboptimal.
iMOD5 defines k northbound, whereas MODFLOW defines it eastbound. We currently resolve this by filling in an angle1 of 90 whenever the model is horizontally anisotropic. This likely leads in overhead from MODFLOW6 having to do a lot of unnecessary trigonometric computations.

Instead, we should rotate the horizontal k's ourselves in iMOD Python.

This after #1208 is resolved, as for that I'd like to keep written MODFLOW6 inputs equal to what iMOD5 writes as much as possible.

[JoerivanEngelen]

@verkaik looked into the MODFLOW6 code and couldn't really see a reason why it would slow computations down a lot. I'm putting this on the sidequest list for now.