[onert] Fix Python typing for the benchmark_inference() function

This commits fixes the typing by allowing passing None as a value for
the input_shapes parameter in the benchmark_inference() function. In
case of input_shapes being None, the function uses the shape retrieved
from the initialized session.

ONE-DCO-1.0-Signed-off-by: Arkadiusz Bokowy <a.bokowy@samsung.com>

Author: arkq

runtime/onert/sample/minimal-python/inference_benchmark.py

@@ -4,7 +4,7 @@
 import numpy as np
 import psutil
 import os
-from typing import List
+from typing import List, Optional
 from onert import infer, tensorinfo
 
 
@@ -47,7 +47,7 @@ def get_validated_input_tensorinfos(sess: infer.session,
     return updated_infos
 
 
-def benchmark_inference(nnpackage_path: str, backends: str, input_shapes: List[List[int]],
+def benchmark_inference(nnpackage_path: str, backends: str, input_shapes: Optional[List[List[int]]],
                         repeat: int):
     mem_before_kb = get_memory_usage_mb() * 1024