Fix copy_to with unsupported variable and constraint attributes

[odow]

Closes #2327

[blegat]

I think it was intentional that we only allow skipping if supports is false because it means the solver does not support it. That allows writing a code that sets starting values and then use it with both Mosek and COSMO.
If there is an error since it would mean that something might have gone wrong and the user should know about it, especially since, last time I checked, if you provide an incomplete dual, SCS will just ignore it so if the user has set all the starting values is pretty confusing.
In the case of the issue you are mentioning, I think it's best to error than to just let the solver ignore all starting values and let the user figure out the hard way that one of its starting values was just ignored. We should at least warn.

[odow]

That's a reasonable argument. I'm okay to close this PR (and the issue) then. We can always wait to see if we get complaints, not that the error message is better.