support saving model defined parameters without add scale_op (#36119)

* add scale_op in model save step is not necessary, just fix the prune method to support static graph and inplace op

* fix jit.save, no need to add scale_op to each outputvar anymore.
fix prune_with_input, now it supports inplace op

* temporarily disable test_trt_dynamic_shape.TRTDynamicShapeOutOfBound2Test

* allow user to export parameters defined in model

python/paddle/fluid/framework.py

@@ -5074,11 +5074,7 @@ class Program(object):
                         else:
                             target_op = op
 
-                if target_op is None:
-                    raise ValueError(
-                        "The target variable used for pruning should have an "
-                        "associated operator that generates it.")
-                else:
+                if target_op is not None:
                     targets_idx.append([target_op.block.idx, target_op.idx])
             else:
                 targets_idx.append([t.block.idx, t.idx])

python/paddle/fluid/io.py

@@ -1426,7 +1426,8 @@ def save_inference_model(dirname,
                 main_program.global_block().create_var(
                     name=target_v.name,
                     shape=target_v.shape,
-                    dtype=target_v.dtype)
+                    dtype=target_v.dtype,
+                    persistable=target_v.persistable)
 
         prepend_feed_ops(main_program, feeded_var_names)
         append_fetch_ops(main_program, fetch_var_names)