Add detection of conflicting global type redefinitions

[tyulpan]

Initially, I ran into the crash described in #1126.
My first idea was to simply ignore duplicate declarations. That would be fine for identical types.

But what if we wanted to do something like this:

global record Widget
  id: integer
end

global record Gadget
  name: string
end

global type Handle = Widget
global type Handle = Gadget

Simply ignoring the duplicate declaration would be a footgun here.

And that's where I ran into another problem. Related to #1032. Currently, redeclaring a global type (record, interface, or type alias) with a different definition isn't reported as an error, and the compiler keeps the first one it finds.

So let's fix that.

Resolves #1126.

[hishamhm]

The one thing I need to check about this PR is: are we sure that adding this does not accidentally introduce structural type compatibility to an otherwise mostly-nominal type system?