[mojo][repl] Fix erroneous destructor call emission in generated REPL…

… source (#35595)

The REPL is doing dirty tricks with variables, turning them into heap
allocations and tracking them in indirect structs. However,
CheckLifetimes
is onto its tricks, which causes it to reject these as incorrect values.

The right solution is to make more invasive changes to the REPL, but
that isn't in the short term plans.  Workaround this by introducing a
horrible hack (tm) op `RefFromPointerUntrackedOp` that creates a
reference that isn't tracked by CheckLifetimes.

This pokes a hole in our nice reference system which makes me very
sad.  A comment saying "don't use this" will prevent other people from
using this for other things ... right???

---------

Co-authored-by: Chris Lattner <clattner@nondot.org>

modular-orig-commit: 7a6b7339522d2657f72d7e8a3a5b51e4d299167f

docs/changelog.md

@@ -37,6 +37,12 @@ and tools. Please add any significant user-visible changes here.
   changed slightly: `mojo build ./test-dir/program.mojo` now outputs an
   executable to the path `./program`, whereas before it would output to the path
   `./test-dir/program`.
+- The REPL no longer allows type level variable declarations to be
+  uninitialized, e.g. it will reject `var s: String`.  This is because it does
+  not do proper lifetime tracking (yet!) across cells, and so such code would
+  lead to a crash.  You can work around this by initializing to a dummy value
+  and overwriting later.  This limitation only applies to top level variables,
+  variables in functions work as they always have.
 
 ### ❌ Removed
 

examples/notebooks/BoolMLIR.ipynb

@@ -105,7 +105,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "var a: OurBool"
+    "fn uninitialized_our_bool():\n",
+    "  var a: OurBool\n"
    ]
   },
   {

examples/notebooks/HelloMojo.ipynb

@@ -65,7 +65,6 @@
         }
       ],
       "source": [
-        "#| XFAIL: *\n",
         "#| CHECK: Hello Mojo!\n",
         "print(\"Hello Mojo!\")"
       ]