Fix inconsistent naming of Tree parameter in Widget trait

[edwloef]

This removes the underscores in front of the argument names in provided Widget methods, in order to make an IDE-autofilled method signature more useful. This was done in a similar manner to https://github.com/Smithay/smithay/blob/bc1d7320f95cdf17f9e7aa6867cccc5903548032/src/xwayland/xwm/mod.rs#L330-L356

It also renames all instances where the Tree argument was named state, as Tree has a state field. This has led to confusion on my part a few times, with the IDE-autofilled method signature clashing with my own naming, in which case I assumed state was the widget's (downcasted) state, not the widget's tree. I can't think of any reasoning as to why it was named tree vs state in different methods, so it makes sense to me to make it consistent everywhere.

[airstrike]

Assuming this change is desirable, would it not be better to favor #[allow(unused_variables)] over let _ = ...?

[edwloef]

I don't really mind either way, I just copied how Smithay dealt with this issue. If #allow's are preferred I can take care of it as well.

[hecrj]

You will rarely need to use all the arguments, so the usefulness is quite subjective.

I prefer to have to opt-in into arguments rather than opt-out.

In any case, this seems like we are adding complexity to the code for better IDE support. Shouldn't the issue be directly addressed in the LSP layer? Maybe a setting in rust-analyzer?

[edwloef]

You will rarely need to use all the arguments, so the usefulness is quite subjective.

You do always need all of them if your widget has child widgets, which to be fair I'm not sure how common that is.

If that's not such a common scenario it would make sense to not remove the underscores, since this is just a small annoyance I noticed while working on a few widgets that had each other as children.

[edwloef]

I've reopened this to only include the fix for #2949 (EDIT: it didn't reopen for some reason? I'll open a new PR)