Add trait impls for refs

[lmondada]

Closes #93

[lmondada]

@aborgna-q What do you think of this solution? Do you see a less verbose way? We'd have to do the same for &mut for both mutable and immutable types...

The better solution would probably to fix auto_impl to support GATs (if that is really the problem), but I don't have time to do that now.

[aborgna-q]

• I'm ok with adding this, but what's your usecase?
• You could use delegate! to reduce the boilerplate. (And also makes sure to add all the #[inline]s).

    delegate! {
        to (*self) {
            fn port_direction(&self, port: impl Into<PortIndex>) -> Option<Direction>;
            fn port_node(&self, port: impl Into<PortIndex>) -> Option<NodeIndex>;
            ...
        }
    }

(see filter.rs)

[lmondada]

I have a matcher trait with the following method

trait Matcher<G: GraphBase> {
    fn find_rooted_matches(&self, G, root: <G as GraphBase>::NodeId);

    fn find_matches(&self, G) where G: petgraph::visit::IntoNodeIndices {
        // some code that requires petgraph::visit::IntoNodeIndices
    }
}

For the implemenation, it's nicer to write

impl<G> Matcher<G> for SinglePatternMatcher<U, (), UnweightedEdge>
where
    G: LinkView + GraphBase<NodeId = NodeIndex> + IntoNodeIndices
{ ... }

rather than

impl<'g, G> Matcher<&'g G> for SinglePatternMatcher<U, (), UnweightedEdge>
where
    G: LinkView
    &G: GraphBase<NodeId = NodeIndex> + IntoNodeIndices
{ ... }

(if that syntax is even valid?)

Does that make sense?

[aborgna-q]

s/IntoNodeIndices/IntoNodeIdentifiers/.

It may be annoying to write if you ever have a generic Matcher, but I think that'll work.

[aborgna-q]

LGTM