Made label data optional for inference and adopted other required changes

[laxmareddyp]

Adapted the architecture preprocessor to conform with the current code guidelines and typical model interface.

• removed call from SegFormerImageSegmenterPreprocessor, perform transformations in the image converter instead
• adapted convert_segformer_checkpoints.py accordingly

These are the points/bugs changed in segformer_image_segmenter_tests.py:

• MiTBackbone parameter names were wrong (had not been adapted for [MiT and SegFormer] Refactor Backbone Arg Names #1958)
• Added label data in train_data for self.run_task_test
• expected_output_shape in test_task was wrong
• use self.assertAllEqual instead of plain assert
• a valid preprocessor needs to be used for self.run_task_test to succeed
• fixed a wrong input shape in test_segformer_call

[sachinprasadhs]

LGTM