Scaladoc: further definition lookup fixes

[abgruszecki]

Lookup supports DFS now, and duplicates more of Scaladoc's old behaviour.

Fixes majority of remaining lookup warnings in stdlib.

[abgruszecki]

@romanowski this PR deals with most of the remaining lookup issues in stdlib. The logic needed to be changed to be closer to original Scaladoc - it was more like Dottydoc before this PR.

We now look up the query in top-level scope to start with, then, if that fails, we look it up in every enclosing scope, including every enclosing package. This seems to be how Scaladoc behaved previously. We also do a DFS when looking up the definitions - for instance, when looking up Array.fill, we first treat Array as a class, fail to look up fill, then retreat to treating Array as object and succeed.

I'll open a PR with associated documentation changes sometime later this week.

[abgruszecki]

FYI right now #11306 and #11305 are, at a glance, the only problems remaining with definition lookup in the doctool.

#11305 could be easily fixed if we could generate a DRI for an inherited definition. From the perspective of definition lookup, it's just a matter of returning two symbols instead of one: one for the definition we looked up, one for the "parent" of the definition, i.e. the symbol through which we looked up the first one.