Support crossreference for imported modules and entities

[robinp]

As final result, the named entries in the import list should be properly hyperlinked.

The module names, like Foo in import Foo (bar, baz), should ref/imports the module node (see #8 ). Also, bar and baz should ref/imports too.

Dotted imports

import Foo (Bar(..)) - should we emit multiple ref/imports edges from the ..s? Or just do nothing there.

Aliased imports

What about aliased imports (regardless of qualification)? For example:

import Data.Text as T
...
foo = T.append "x" "y"

Easiest is we do nothing. In any case, T.append will still refer to the same node. If we emitted a defines/binding on T of the import line, then the T of T.append might ref that. But there's no really suitable node in the kythe schema (unless we shoehorn talias + aliases), and the benefit is dubious. Also, the def/binding would need to come from an implicit anchor, since multiple modules can be aliased to the same name (so the T in the import line would also be a reference).

[robinp]

Info: GHC AST stores these as ImportDecl available from the RenamedSource.

Also, the implicit prelude import happens from a funny location. If the import is implicit, we shouldn't attach a Span to it in the translate layer rather (and then the reference will have an implicit anchor in Kythe).

[robinp]

Module names are referenced now, thanks to Ivan.

To be done: to keep it simple, ref/imports the explicitly mentioned imported names, maybe ref/imports/hiding those in a hiding clause.

[robinp]

Actually, both the module name and the import statements seem to ref the module entry, but noone is defines/binding it. So Kythe navigation to definition doesn't work.

[robinp]

Actually now the module name just refs the node, but doesn't emit a defines/binding, so Kythe navigation to definition doesn't work on imports.