Restore error when unknowns in the initialization system don't have guesses

[hersle]

Another attempt to fix #2873 in a simpler way; supersedes #2874.

Checklist

• Appropriate tests were added
• Any code changes were done in a way that does not break public API
• All documentation related to code changes were updated
• The new code follows the
  contributor guidelines, in particular the SciML Style Guide and
  COLPRAC.
• Any new documentation only uses public API

[hersle]

I don't understand what's going on with the DDE test.

Seems weird that just disabling one early-return optimization causes one test to fail. Looks like there is just a t that is not substituted. Is there something DDE-specific going on here, or have DDEs just not gotten the non-autonomous treatment?

[ChrisRackauckas]

Looks like there is just a t that is not substituted. Is there something DDE-specific going on here, or have DDEs just not gotten the non-autonomous treatment?

DDEs naturally do a bit more with t since it's common to have say x(t - tau), or x(t - y(t)). I wonder if those cases trip it up.

[ChrisRackauckas]

This looks good to go though? Just the latexify regression test needs an update, but I'd be fine handling that after merge.

[hersle]

Yes, that's what I realized. It seems that DDEs relied on the early return, which is why they pass tofloat = false in the first place.

I think it's good to merge, yes.

[ChrisRackauckas]

Thanks for seeing this one through. I hadn't found the time to do it since it was pretty non-trivial, but in the end I think you figured out a good solution.