make FieldName required in the CST

[OmarTawfik]

Making this a required property for all nodes (replacing the Option<T>) makes it much easier to handle/match in client code.

• Use FieldName::Item and FieldName::Separator for SeparatedItem children.
• Use FieldName::Item for RepeatedItem children.
• Use FieldName::LeadingTrivia and FieldName::TrailingTrivia for children of LeadingTrivia and TrailingTrivia.
• Are there any other cases?

[AntonyBlakey]

The ability to have un-named nodes is critical to the query system!

[Xanewok]

I'm also in the camp of having the names optional. If it's all about Option<T> handling, we can surely inline the None variant directly in the FieldName but it makes sense not to name every node there is.

[OmarTawfik]

I really think we should make it required everywhere. From an API perspective, if i'm iterating on children, or even filtering/inspecting a specific child, I should be able to read/know its name, and it should be strongly typed, and finite/matchable. All such is provided by the FieldName enum. But using Option<FieldName> and None force me to then rely on my own knowledge of how the parser works to know what to do next.

Re: the query system: I think the fact that something has a name or not shouldn't interfere with the querying system, as they are independent, and can actually be confusing in some cases, like the separator: Comma that separates TupleDeconstructionElements with empty members. I think it should derive its contentful-ness from other existing properties accordingly.

But we also discussed the cost of adding a side channel to record such information, and it is not clear whether it is worth doing now.

[OmarTawfik]

Another idea: we might not need a side channel after all, if we can derive this info from existing node kinds (example below). We would need to use #808 to see if this affects query engine performance:

fn is_node_contentful(parent_kind: &RuleKind, child_kind: &TokenKind) -> bool {
    match (parent_kind, child_kind) {
        // for `StructItem`, add children that appear in `terminated_by`/`delimited_by`:
        (RuleKind::Statement, TokenKind::Semicolon) => false,
        (RuleKind::Block, TokenKind::LeftCurly | TokenKind::RightCurly) => false,
        // .....

        // For `Separated`, add the `separator`:
        (RuleKind::Arguments, TokenKind::Comma) => false,
        (RuleKind::Parameters, TokenKind::Comma) => false,
        // .....

        // Also trivia:
        (_, TokenKind::Whitespace | TokenKind::Newline | TokenKind::Comment) => false,

        // otherwise, contentful:
        _ => true,
    }
}

[OmarTawfik]

Let's defer this until we have more experience/feedback on the API.