Extend backdating checks to imports and globals

[Keno]

Previously, backdated binding warnings were only emitted for constants. This commit extends the backdating mechanism to handle imports and globals. This change was requested by triage to improve compatibility of the new binding partition mechanism given issues like #58511.

With this change the behavior is as follows:

julia> function foo()
           include_string(Main, "a = 42")
           return a
       end

julia> foo()
WARNING: Detected access to binding `Main.a` in a world prior to its definition world.
  Julia 1.12 has introduced more strict world age semantics for global bindings.
  !!! This code may malfunction under Revise.
  !!! This code will error in future versions of Julia.
Hint: Add an appropriate `invokelatest` around the access to this binding.
To make this warning an error, and hence obtain a stack trace, use `julia --depwarn=error`.
42

With julia --depwarn=error, the access is rejected:

julia> foo()
ERROR: UndefVarError: `a` not defined in `Main`

I am still not a huge fan of the backdating mechanism in general, but I think this is the least objectionable of the alternatives and provides a big warning to let people know that something is wrong.

Largely written by Claude.

[KristofferC]

I pretty much think this is too risky to put in this late in the release process, and that the current status quo is the lesser evil.

For example, this PR fails CI on a random assortment of tests, and Revise stops working:

┌ Error: Failed to revise /tmp/SUk8MQkftz/Order1/src/file1.jl
│   exception =
│    UndefVarError: `Ord1` not defined in `Order1`
│    Suggestion: add an appropriate import or assignment. This global was declared but not assigned.
│    Stacktrace:
│     [1] top-level scope
│       @ /tmp/SUk8MQkftz/Order1/src/file1.jl:1
│    Revise evaluation error at /tmp/SUk8MQkftz/Order1/src/file1.jl:1
│ 
└ @ Revise ~/.julia/packages/Revise/JK6mi/src/packagedef.jl:741

[Keno]

Rebinding test failure is #59613, so we should merge that first.

[Keno]

effects.jl is the combination of a pre-existing codegen model bug (will PR separately) with #59613

[Keno]

will PR separately

#59766

[vtjnash]

Please tell Claude its implementation of this particular test is rubbish and to redo it correctly. Approximately:

diff --git a/test/worlds.jl b/test/worlds.jl
index 6dcef169fba..b69f92d2d01 100644
--- a/test/worlds.jl
+++ b/test/worlds.jl
-# Test that implicit imports are NOT backdated
+# Test that conflicting imports are still detected and handled
 module BackdateNoImplicitTest
     using Test
     module Source1
@@ -693,7 +689,22 @@ module BackdateNoImplicitTest
         shared_name() = 2
     end
     const world_before = Base.get_world_counter()
-    using .Source1, .Source2
-    @test_throws UndefVarError shared_name()
+    @test_throws UndefVarError shared_name
+    using .Source1
+    @test shared_name === Source1.shared_name
+    @test shared_name() === 1
+    if Base.JLOptions().depwarn <= 1
+        @test Source1.shared_name === @test_warn "Detected access to binding `BackdateNoImplicitTest.shared_name`" Base.invoke_in_world(world_before, getglobal, @__MODULE__, :shared_name)
+    else
+        @test_throws UndefVarError Base.invoke_in_world(world_before, getglobal, @__MODULE__, :shared_name)
+    end
+    @test isdefinedglobal(@__MODULE__, :shared_name)
+    @test !Base.invoke_in_world(world_before, isdefinedglobal, @__MODULE__, :shared_name)
+
+    using.Source2
+    @test_throws UndefVarError shared_name
+    @test !isdefinedglobal(@__MODULE__, :shared_name)
+    @test_throws UndefVarError Base.invoke_in_world(world_before, getglobal, @__MODULE__, :shared_name)
+    @test !isdefinedglobal(@__MODULE__, :shared_name)
     @test !Base.invoke_in_world(world_before, isdefinedglobal, @__MODULE__, :shared_name)
 end

[Keno]

I think this ship has probably sailed at this point.