[FIX] account_reconcile_model_oca: Fix reco model matching on exact_token

When the payment label is a single word, we try to match the full value and not only the alphanumeric values.
However, the matching was made on the token that was already sanitized using the [0-9a-zA-Z\s] regex.
Instead, let's match the full value.

Related to bccc0ae0e8

account_reconcile_model_oca/models/account_reconcile_model.py

@@ -369,7 +369,7 @@ class AccountReconcileModel(models.Model):
 
             # Exact tokens.
             if len(tokens) == 1:
-                exact_tokens.append(tokens[0])
+                exact_tokens.append(text_value)
         return numerical_tokens, exact_tokens, text_tokens
 
     def _get_invoice_matching_amls_candidates(self, st_line, partner):

account_reconcile_model_oca/tests/test_reconciliation_match.py

@@ -1433,6 +1433,17 @@ class TestReconciliationMatchingRules(AccountTestInvoicingCommon):
                 {},
             )
 
+        with rollback():
+            # Test Matching on exact_token.
+            term_line.name = "PAY-123"
+            st_line.payment_ref = "PAY-123"
+
+            # Matching if no checkbox checked.
+            self.assertDictEqual(
+                rule._apply_rules(st_line, None),
+                {"amls": term_line, "model": rule},
+            )
+
         with self.subTest(
             rule_field="match_text_location_label", st_line_field="payment_ref"
         ):