Fix memory_used OID parsing for Mikrotik devices

Handle OctetString values returned as bytes by converting them to int
using little-endian byte order (4 bytes) for Mikrotik devices.

snmp2mqtt.py

@@ -278,10 +278,29 @@ async def get_snmp(req):
                 continue
             else:
                 for varBind in varBinds:
-                    logging.debug(f"{req['device_name']} {oid['name']} => {oid['type'](varBind[1])}")
+                    raw_value = varBind[1]
+                    
+                    # Handle different SNMP value types
+                    if oid['type'] == int and hasattr(raw_value, '__bytes__'):
+                        # Convert OctetString/bytes to int
+                        # For Mikrotik devices, memory values might be in the first 4 bytes
+                        raw_bytes = bytes(raw_value)
+                        # Try to interpret as integer, handling different byte orders
+                        try:
+                            if len(raw_bytes) >= 4:
+                                # Try little-endian first (common for Mikrotik)
+                                value = int.from_bytes(raw_bytes[:4], byteorder='little', signed=False)
+                            else:
+                                value = int.from_bytes(raw_bytes, byteorder='big', signed=False)
+                        except (ValueError, TypeError):
+                            value = raw_value
+                    else:
+                        value = raw_value
+                    
+                    logging.debug(f"{req['device_name']} {oid['name']} => {oid['type'](value)}")
                     
                     # Cast to the right type
-                    value = oid['type'](varBind[1])
+                    value = oid['type'](value)
                     
                     # Apply operation if defined
                     if 'operation' in oid: