Implement alternate and recommended way to set properties of abstract base class

See: https://github.com/lammps/lammps/security/code-scanning/476

python/lammps/mliap/mliap_unified_abc.py

@@ -4,12 +4,13 @@ import pickle
 class MLIAPUnified(ABC):
     """Abstract base class for MLIAPUnified."""
 
-    def __init__(self):
-        self.interface = None
-        self.element_types = None
-        self.ndescriptors = None
-        self.nparams = None
-        self.rcutfac = None
+    def __init__(self, interface = None, element_types = None,
+                 ndescriptors = None, nparams = None, rcutfac = None):
+        self.interface = interface
+        self.element_types = element_types
+        self.ndescriptors = ndescriptors
+        self.nparams = nparams
+        self.rcutfac = rcutfac
 
     @abstractmethod
     def compute_gradients(self, data):

python/lammps/mliap/mliap_unified_lj.py

@@ -6,16 +6,13 @@ class MLIAPUnifiedLJ(MLIAPUnified):
     """Test implementation for MLIAPUnified."""
 
     def __init__(self, element_types, epsilon=1.0, sigma=1.0, rcutfac=1.25):
-        super().__init__()
-        self.element_types = element_types
-        self.ndescriptors = 1
-        self.nparams = 3
+        # ARGS: interface, element_types, ndescriptors, nparams, rcutfac
+        super().__init__(None, element_types, 1, 3, rcutfac)
         # Mimicking the LJ pair-style:
         # pair_style lj/cut 2.5
         # pair_coeff * * 1 1
         self.epsilon = epsilon
         self.sigma = sigma
-        self.rcutfac = rcutfac
 
     def compute_gradients(self, data):
         """Test compute_gradients."""