Support paths when defining supertrait for [com_interface]

[Rantanen]

Instead of having to use the IUnknown everywhere, this PR adds support for:

#[com_interface(...)]
trait Foo: com::interfaces::iunknown::IUnknown {}

Most of the bits were very straight forward. The one places where the implementation got a bit complex (and a bit fragile) is the #[derive(VTable)] expansion, which used to derive the interface ident from the field name (making assumptions such as all interfaces starting with I) - the new implementation assumes the type is defined as an associated type of the original interface:

  <dyn path::to::some::Interface as ...>::VTable
// ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ use this

This branch is in a conflict with #108 and will fail to build with #107. I'll merge things around once I have a better idea what PRs will land and in which order.

[rylev]

Looks great. Lets add another test that makes sure use ... works. Perhaps we could have two modules one where use ... is used and one where it is not and then we can test that both work.

[Rantanen]

Added test for use and relative path with super::. Was debating whether to write these in a different file or not, but given the tests are so closely related I just put them in the same file.

.. technically IUnknown is still a Path. :)

[Rantanen]

Also rebased on top of the new master branch to resolve the conflicts with the expect branch.

[rylev]

Looks great!