have a nice function to create a default fallback -- and warn if it doesnt have enough properties

[berndbischl]

marc added some code in encapsulate to preconfigure a fallback.

maybe that should simply be a helper. so it can be called from the outside.

i also think: on setting the fallback you PROBABLY wanna warn the user if you fallback can do "less", e.g. doesn't have the same predict_type (error?) or can do no weights....

we don't have to be SUPER precise here or "block" actions. but at least warn about obvious problems?

[sebffischer]

this should maybe take into account that graphlearners have all available properties (it junk this is a workaround because inferring these properties is kind of tricky)

[be-marc]

1151