Things I'm for in this MR:

• Adding stage checks to the functions. I think there is no better way to do this than what you've done.
• The transform functionality. What is the difference between this and calling presolve however? I would need to this distinction made clear.

Things I don't like:

• Why do users need to add basic linear constraint? Shouldn't this be done with _createConsLinear? The basic include is supposed to save users implementing certain callbacks. This is easy in Python though and doesn't take up much code, and I think it's more confusing to provide this functionality.

Hey @Opt-Mucca, thanks for looking into this.

After transformProb the problem is in stage TRANSFORMED, so presolving wasn't called yet. I'm not entirely sure what presolving does after the problem is transformed though, but agreed that we should make it clear.

The idea for adding _createConsBasicLinear was because it might help out in other projects. There is this ML4LB repo that asks users to install PySCIPOpt and then compile their own version with these added methods. I'm not entirely sure that this is actually needed (they use it here), but it's still an annoyance for the users, as described in #910.

To be fair, in the documentation we do have this line, but we might want to remove it:

Ideally speaking, we want every SCIP function to be wrapped in PySCIPOpt.