This also heavily refactors our existing scanner
infrastructure to accomodate both Antlr based lexers
and LSP-based semantic token providers.
As the latter also support supplying token modifiers,
support for this has been added as well
(e.g., rendering static things in italics).

This can be used to override equality checks to
use reference equality, which can be useful for
e.g. records.

This also improves the tooltip a bit,
by giving it both a minimum and maximum width.

This bug previously occurred if the entries' titles
contained any slashes, as this would induce different
behavior in the Transform.Find method.

This displays a code window for a given path and range,
instead of for a given graph element.
It works by calling the `FittingElements` extension method
on the graph, also implemented in this commit.

This, in turn, finds graph elements matching
the given path and range, ordered by how well
they fit.

To make things easier and less computationally intensive,
we now simply make the line number of the marked line red
instead of marking the whole line with some background color.

This can be triggered by right-clicking any element
in the code window.

This gives us a much easier option of using existing
icons, rather than having to import an image
and convert it to a sprite each time.
It also improves performance a bit.

… into lsp-for-code-windows