ML-484 add Keras Applications support

[xwang2713]

Reference this link for detail: https://keras.io/api/applications/

Example usage:

modelName := 'MobileNetV2';
fdef := '''input_shape=(224,224,3),include_top=True,weights=None,classes=2''';
compileDef := '''compile(optimizer=tf.keras.optimizers.Adam(),
                 loss=tf.keras.losses.binary_crossentropy,
                 metrics=[tf.keras.metrics.BinaryAccuracy()])
                 ''';
s := GNNI.GetSession();
mod := GNNI.DefineKAModel(s, modelName, fdef, compileDef);

Following models work:
MobileNetV2';
Xception
NASNetMobile
DenseNet121
MobilNetV2

Known failed ones:

1. Dataset too large to output to workunit
   VGG19
   ResNet152V2
   InceptionResNetV2
2. 'numpy.float64' object cannot be interpreted as an integer
   EfficientNet series, for example 'EfficientNetB0'
   https://track.hpccsystems.com/browse/ML-497

[xwang2713]

@RogerDev @lily please review

[xwang2713]

@RogerDev @lilyclemson please review this or maybe we can defer and close this.