model.load_model now supports loading of concatenate layers.
python tensorflow exports concatenate layers in an extra nested array in the manifest so added a check for that in generic_utils.cs.
Concatenate was missing the build=true, this fix prevents the layer being build multiple times.
Concatenate has 2 or more input nodes so List<NodeConfig> was required instead of just NodeConfig in Functional.FromConfig.cs.
Added missing axis JsonProperty attribute for MergeArgs (used by Concatenate)

观察到的现象是，一些模型增大batchsize后，会在首个epoch的中途爆显存不足，只要过了一个epoch后，就能完整训练。同样的batchsize在python下能设置大得多的值。
最后使用最小训练代码分析出，是每个step之后，图片加载到显存里的数据没有释放导致的。
在寻找释放显存接口没有结果的时候，直接使用了GC.Collect();可以让显存主动回收。
因此当前的修复方案是在每个step里，都执行一次 GC.Collect(); 用来释放显存资源。

the bool to tensor has a bug, if in init the training is False, the program not start.

Fix my bad:

Bad handling between test_function and test_step_multi_inputs_function.

IModel.Dictionary<string, float> evaluate(NDArray, NDArray, ...) is now IModel.Dictionary<string, float> evaluate(Tensor, Tensor, ...)
Merge Model.Evaluate.test_step_multi_inputs_function(...) and Model.Evaluate.test_function(...)

Note: An internal function need to add an explicit cast in Tensor

Fix issue if data_handler.steps() > 1

* Add CheckpointReader and corresponding C APIs.

* Add essential components of SavedModel format loading.

* Add checkpoint reading for SavedModel format loading.

* Revise customized json converters.

* Add support for loading models from python.

* Fix the duplicated weights in Keras.Model.

* Add alexnet loading test and check for loaded weights.

* Fix ci error caused by branch merge.

* Resolve the comments and errors.

* Fix the stucking of training when loading model.

* Fix the stucking of training when loading model.

* fix intptr.

---------

Co-authored-by: Haiping Chen <haiping008@gmail.com>