Remove emin=1-emax restriction and specify emin explicitly

[mmikaitis]

At present a test in ctest is failing, while matlab/octave tests pass.

[mfasi]

Thanks, Mantas – this is an impressive PR!

I think it's essentially ready to be merged into main – I have just a few minor comments.

My main question is whether the order of $e_\min$ and $e_\max$ makes a difference. Currently, we have $e_\max$ first, and this is done consistently in documentation and code.

In books and documentation, $e_\min$ often comes first, possibly because $e_\min \le e_\max$. One notable exception is the IEEE standard, but this makes sense to me: there $e_\min$ is defined in terms of $e_\max$, so this is the only correct order.

Should we swap them or keep them as they are?

[mfasi]

Maybe we should state explicitly the version of the two packages where things started to diverge?

[mmikaitis]

I added a list of items. Feel free to add further items or modify.

[mfasi]

There are a few other differences though: the random number generator can be passed to chop, the OCP formats are (will be) different, etc. Do we want to specify what differences are there?

[mmikaitis]

Now linking to the list of differences in README.md

[mfasi]

OK. I think this is true for Windows. Were you able to compile the PCG on a Mac and link it to the MEX interface? If yes, then I'll explicitly mention that this is true for Windows only; otherwise, I'll say it's true unless you're on Linux.