Tfcoreml

Changes:

- Add support for TensorFlow 2.x file format (.h5, SavedModel, and concrete functions).
- Add support for several new ops, such as `AddV2`, `FusedBatchNormV3` for iOS 13+.
- Bug fixes in the TensorFlow converter's op fusion graph pass.
- New `minimum_ios_deployment_target` flag (replacing previous `target_ios` which causes confusions). This flag can be utilized for setting **minimum** targeting iOS, e.g., the model converted with `minimum_ios_deployment_target='12'` will support iOS 12, iOS 13, and later.

Known Issues:

- `tf.keras` model conversion supported only with TensorFlow 2
- Currently, there are issues while invoking the TensorFlow 2.x model conversion in Python 2.x.
- Currently, there are issues while converting `tf.keras` graphs that contain recurrent layers.

This release was focused on adding many new layers and operators which can be found [here](https://github.com/apple/coremltools/blob/2a08445ade3c0da81fb2b25cf6de9f88c993be0c/coremltools/converters/nnssa/coreml/ssa_converter.pyL330). The coremltools release 3.0 can be found [here](https://github.com/apple/coremltools/releases).

[Examples for model conversions and how to use control flows](https://github.com/tf-coreml/tf-coreml/tree/master/examples)

* Added the argument `target_ios` to choose the Core ML spec version that is produced by the converter. `target_ios='13'` will enable the converter to use all the new layers added in Core ML 3.
* Added `custom_conversion_function` option where users can provide their own custom conversion function similar to these [examples](https://github.com/apple/coremltools/blob/master/coremltools/converters/tensorflow/test/test_custom_layer.py) for the custom conversion use case.

Supports new model specification version 4 with target iOS 13 or later, the Core ML model specification can be found in [proto](https://github.com/apple/coremltools/blob/c6e7d15e3aef676a60247fea235da58aedbfcfd7/mlmodel/format/NeuralNetwork.protoL535) or [documentation](https://apple.github.io/coremltools/coremlspecification/sections/NeuralNetwork.html).

Any questions or concerns related to this release can be submitted as an issue and will be review by the team. All comments are welcomed and will be used to improve the existing documentation.

* added support for resizeBilinear and cropAndResize ops
* added support for 8-bit QuantizedConv2D, QuantizedMatMul ops
* some bug fixes

- Adds Python 3 support
- Adds custom layer support
- Miscellaneous bug fixes