Rollup merge of #124444 - compiler-errors:eval, r=lcnr

Record certainty of `evaluate_added_goals_and_make_canonical_response` call in candidate

Naming subject to bikeshedding, but I will need this when moving `select` to a proof tree visitor.

r? lcnr

compiler/rustc_middle/src/traits/solve/inspect.rs

@@ -122,6 +122,12 @@ pub enum ProbeStep<'tcx> {
     /// used whenever there are multiple candidates to prove the
     /// current goalby .
     NestedProbe(Probe<'tcx>),
+    /// A call to `EvalCtxt::evaluate_added_goals_make_canonical_response` with
+    /// `Certainty` was made. This is the certainty passed in, so it's not unified
+    /// with the certainty of the `try_evaluate_added_goals` that is done within;
+    /// if it's `Certainty::Yes`, then we can trust that the candidate is "finished"
+    /// and we didn't force ambiguity for some reason.
+    MakeCanonicalResponse { shallow_certainty: Certainty },
 }
 
 /// What kind of probe we're in. In case the probe represents a candidate, or

compiler/rustc_middle/src/traits/solve/inspect/format.rs

@@ -132,6 +132,9 @@ impl<'a, 'b> ProofTreeFormatter<'a, 'b> {
                     }
                     ProbeStep::EvaluateGoals(eval) => this.format_added_goals_evaluation(eval)?,
                     ProbeStep::NestedProbe(probe) => this.format_probe(probe)?,
+                    ProbeStep::MakeCanonicalResponse { shallow_certainty } => {
+                        writeln!(this.f, "EVALUATE GOALS AND MAKE RESPONSE: {shallow_certainty:?}")?
+                    }
                 }
             }
             Ok(())