[Example] support more TFLite image classification models

[huningxin]

The existing example (https://github.com/intel/webml-polyfill/tree/master/examples/mobilenet) only supports TFLite MobileNet model for image classification. This task is to make the example more general to support more TFLite image classification models hosted here:

https://github.com/tensorflow/tensorflow/blob/master/tensorflow/contrib/lite/g3doc/models.md#image-classification-float-models

It requires:

1. refactor the example to load, compile and execute multiple models
2. identify the unsupported ops if the model is not supported

[Wenzhao-Xiang]

These are the models that are supported with present ops:
Mobilenet_V1 / Mobilenet_V2 / Inception_V3 / Squeezenet

And these are the models that are unsupported now and the ops they need:

model	FullyConnected	Mean	Pad	Split	StridedSlice
Densenet	√	√
Inception_V4	√
Inception_Resnet_V2	√
Nasnet_Mobile	√	√	√	√	√
Nasnet_Large	√	√	√	√	√
ResNet_V2_50		√	√
ResNet_V2_101		√	√
MnasNet_1.0_224	√	√

Besides, Mobilenet_V1 need the true Squeeze operator, now we use Reshape to replace it.

[Wenzhao-Xiang]

NNAPI now support FullyConnected / Mean / Pad / StridedSlice.
And this is the definition of Split in tensorflow.js.

[Wenzhao-Xiang]

This is the tflite model for Image Classification and the ops they include.

Model \ Op Type	ADD	MUL	CONV_2D	DEPTHWISE_CONV_2D	AVERAGE_POOL_2D	MAX_POOL_2D	CONCATENATION	RESHAPE	SOFTMAX	SQUEEZE	FULLY_CONNECTED	Mean	PAD	SPLIT	STRIDED_SLICE	RELU
Mobilenet_V1			√	√	√				√	√
Mobilenet_V2	√		√	√	√			√	√
SqueezeNet			√		√	√	√	√	√
Densenet	√	√	√		√	√	√		√		√	√
Inception_V3			√		√	√	√	√	√
Inception_V4			√		√	√	√		√		√
Inception_Resnet_V2	√		√		√	√	√				√
Nasnet_Mobile	√	√	√	√	√	√	√		√		√	√	√	√	√	√
Nasnet_Large	√	√	√	√	√	√	√		√		√	√	√	√	√	√
ResNet_V2_50	√	√	√			√		√	√			√	√
ResNet_V2_101	√	√	√			√		√	√			√	√
MnasNet_1.0_224	√		√	√							√	√

And Here is the Wiki page