Object Detection (tf2) use of BestExporter

Prerequisites

Please answer the following question for yourself before submitting an issue.

• [ X] I checked to make sure that this feature has not been requested already.

1. The entire URL of the file you are using

https://github.com/tensorflow/models/blob/master/research/object_detection/model_lib_v2.py

2. Describe the feature you request

When training for example an efficientdet model for object detection. Tf OD API just keeps latest checkpoints, and maybe, if you let the model training for a long time, latest checkpoint are overfitted models. I have seen examples of BestExporter class to keep N best models in eval, but these examples are for TF1 (using BestExporter in EvalSpec class).
I think there should be a way to use BestExporter of another method to keep best models during training

3. Additional context

Add any other context about the feature request here.

4. Are you willing to contribute it? (Yes or No)

Yes