[bugfix?] Avoid forgetting theory conflict clauses

When theory raises a conflict, it is analysed, and the backtracck clause
that result is added to the solver, however, I didn't find yet a
satisfying answer as to wether the original clause is implied (or not)
by this backtrack clause, so in order not to lose information, we also
add the original conflict clause when it comes from the theory (because
if not, then it comes from a conflict detected during propgation, so the
conflict clause is actually already attached).

src/core/internal.ml

@@ -1006,7 +1006,14 @@ module Make
       match propagate () with
       | Some confl -> (* Conflict *)
         incr conflictC;
-        add_boolean_conflict confl
+        (* When the theory has raised Unsat, add_boolean_conflict
+           might 'forget' the initial conflict clause, and only add the
+           analyzed backtrack clause. So in those case, we use add_clause
+           to make sure the initial conflict clause is also added. *)
+        if confl.attached then
+          add_boolean_conflict confl
+        else
+          add_clause confl
 
       | None -> (* No Conflict *)
         assert (env.elt_head = Vec.size env.elt_queue);