[red-knot] remove Type::Unbound

[carljm]

We should not represent "unbound" as a type, because it's only meaningful within a scope, it shouldn't "travel" across scopes as a type.

Instead, we should explicitly use the use-def map methods to check whether a name may be unbound in the places where it matters.

[MichaReiser]

Depends on #13563

[MichaReiser]

Some extra notes from my conversation with @carljm the other day regarding on how this should be implemented

In general, when we query the type of a symbol from another scope, and that symbol either doesn't exist or is definitely unbound, that should be a diagnostic and we should return Unknown.
if the symbol is conditionally bound, we should just return the type from the bound paths (and possibly a diagnostic in this case as well)
the "diagnostic" part may require some of these methods to gain a more complex Outcome-style return type
but it would also be possible to remove Unbound (and just return Unknown in those cases) without handling the diagnostic part at all yet