Const suggestion check: an ambiguity reason is not re-evaluated after the fixpoint (not reproduced) #248

Follow-up of #217 and #239 (final review, M2). Not reproduced.

In src/analysis/ConstSuggestionMatch.cpp, the correspondence steps 3 to 5 share one fixpoint. When a step meets an ambiguity (one symbol that becomes two, two that become one, or no one-to-one counterpart through the template arguments), it records the reason in missing the first time and goes on. The reason is never re-evaluated after the fixpoint: if a later round pairs the symbols that made the case ambiguous, the recorded reason may describe a state that no longer holds.

Not reproduced: no input has been found where a later round resolves an ambiguity already recorded, and none of the suite, the T tests or the corpora shows it. The concern comes from reading the code.

To do: either find an input that shows it, or show that a recorded ambiguity cannot be resolved later. If it can, recompute the reason from the final pairs.