lowering: increment world age after toplevel expressions (#59703)

When a macro expands to `:toplevel` expression, it seems reasonable (as
would be if the expression were not wrapped in `:toplevel`) that any
contained struct definitions be available thereafter. For example, this
applies to `@enum`, which countrly causes world age erros when
attempting to define a method on an enum within the same expression. Fix
this by having lowering insert and explicit `latestworld` after
toplevel.

Fixes #59429

Written by Claude

(cherry picked from commit 0c8768f)

Author: Keno

src/julia-syntax.scm

@@ -5032,6 +5032,7 @@ f(x) = yt(x)
              (check-top-level e)
              (let ((val (make-ssavalue)))
                (emit `(= ,val ,e))
+               (emit `(latestworld))
                (if tail (emit-return tail val))
                val))
 

test/worlds.jl

@@ -589,3 +589,21 @@ module C57316; import ..X57316.Y57316 as Z, .Z.Y57316 as W; end
 @test !isdefined(B57316, :X57316)
 @test !isdefined(C57316, :X57316)
 @test !isdefined(C57316, :Y57316)
+
+# issue #59429 - world age semantics with toplevel in macros
+module M59429
+using Test
+macro new_enum(T::Symbol, args...)
+   esc(quote
+      @enum $T $(args...)
+      function Base.hash(x::$T, h::UInt)
+        rand(UInt)
+      end
+    end)
+end
+
+@new_enum Foo59429 bar59429 baz59429
+
+# Test that the hash function works without world age issues
+@test hash(bar59429, UInt(0)) isa UInt
+end