[webml/NNAPI] Add Error Handling

[Christywl]

There is no error handling. For example, passing an invalid parameter should throw an error. Error handling should be defined. More invalid scenes: addOperation():

• One of the inputs is not the required type(TENSOR_FLOAT32 or TENSOR_QUANT8_ASYMM)
• The input or output tensor rank is not matched with the required
• The type of the second tensor is not same with the first one for ‘ADD’ operation
• The length of inputs are not 7 or 10 for "AVERAGE_POOL_2D" operation
• The input tensors don't have identical type or the same dimensions for "CONCATENATION" operation
• The value of fuse code is invalid for ADD, AVERAGE_POOL_2D, CONV_2D, DEPTHWISE_CONV_2D, MAX_POOL_2D and MUL operations ...

identifyInputsAndOutputs():

• The output is equal to the input
• target output or input wasn't previously added
• Pass more than two parameters

finish(): pass a parameter createCompilation(): call this function when the model is not being finished ...

How to Reproduce:

1. Launch chromium and enable WebML flag
2. Visit https://brucedai.github.io/nt/test/index-local.html?backend=nnapi

[canxcao]

Test case refer to Link (#80, #81).