Consistency checker doesn't walk the tree (for mixins)

[Elchi3]

I'm working on demixing!

One thing I noticed is that the consistency checker apparently doesn't walk the tree.

Assume you have api/ShadowRoot.json
Basic support (root __compat object, so api.ShadowRoot) says chrome 53.

Now you add to that api.ShadowRoot tree in a file called _mixins/DocumentOrShadowRoot__ShadowRoot.json
And so the tree gets api.ShadowRoot.activeElement and it says chrome 1.

This makes no sense. ShadowRoot sub features can't exist before the ShadowRoot main interface exist. The consistency lint doesn't detect this.

[ddbeck]

I don't have a precise solution in mind for this yet, but I think this mostly stems from the fact that our checks run on a per-file basis. We might use this as an opportunity to distinguish between checks on the source (i.e., that it parses and it's formatted correctly), checks on a single feature (or its components—such as note formatting, etc.), and checks against a tree or subtree (i.e., that a tree in a given file or the whole tree is consistent).

[foolip]

This is a bit of an issue for #10646, where I know there's some lack of consistency introduced, but I don't know how much.

[queengooborg]

This was resolved by introducing the next-gen linter in #16437!