Normalize generic parameters for LoginSecondMessage

(aka use the CipherSuite just like the others)

Repair the generics in KE trait: it makes sense that generate_ke[1-3] should operate on the same key representation.

src/tests/opaque_ke_test.rs

@@ -471,19 +471,15 @@ fn test_l3() -> Result<(), PakeError> {
     let parameters = populate_test_vectors(&serde_json::from_str(TEST_VECTOR).unwrap());
 
     let mut client_e_sk_rng = CycleRng::new(parameters.client_e_sk.to_vec());
-    let (l3, shared_secret, export_key_login) = ClientLogin::<X255193dhNoSlowHash>::try_from(
-        &parameters.client_login_state[..],
-    )
-    .unwrap()
-    .finish(
-        LoginSecondMessage::<EdwardsPoint, X25519KeyPair, TripleDH, sha2::Sha256>::try_from(
-            &parameters.l2[..],
-        )
-        .unwrap(),
-        &Key::try_from(&parameters.server_s_pk[..])?,
-        &mut client_e_sk_rng,
-    )
-    .unwrap();
+    let (l3, shared_secret, export_key_login) =
+        ClientLogin::<X255193dhNoSlowHash>::try_from(&parameters.client_login_state[..])
+            .unwrap()
+            .finish(
+                LoginSecondMessage::<X255193dhNoSlowHash>::try_from(&parameters.l2[..]).unwrap(),
+                &Key::try_from(&parameters.server_s_pk[..])?,
+                &mut client_e_sk_rng,
+            )
+            .unwrap();
 
     assert_eq!(
         hex::encode(&parameters.shared_secret),