Add missing standardize_image key to metadata

[jppgks]

A Ludwig model from before #2408 failed to preprocess_for_prediction:

File /workspaces/ludwig/ludwig/features/image_feature.py:479, in ImageFeatureMixin._finalize_preprocessing_parameters(preprocessing_parameters, encoder_type, column)
    475 assert isinstance(num_channels, int), ValueError("Number of image channels needs to be an integer")
    477 average_file_size = np.mean(sample_num_bytes) if sample_num_bytes else None
--> 479 standardize_image = preprocessing_parameters["standardize_image"]
    480 if standardize_image == "imagenet1k" and num_channels != 3:
    481     warnings.warn(
    482         f"'standardize_image=imagenet1k' is defined only for 'num_channels=3' but "
    483         f"detected 'num_channels={num_channels}'.  For this situation setting 'standardize_image=None'.",
    484         RuntimeWarning,
    485     )

KeyError: 'standardize_image'

[github-actions]

Unit Test Results

         6 files  ±  0           6 suites  ±0   4h 50m 42s ⏱️ - 8m 35s
  3 892 tests ±  0    3 817 ✔️ ±  0    74 💤 ±  0  1 ❌ ±0 
11 532 runs   - 95  11 327 ✔️  - 78  204 💤  - 17  1 ❌ ±0 

For more details on these failures, see this check.

Results for commit 42efd48. ± Comparison against base commit d83693a.

♻️ This comment has been updated with latest results.

[tgaddair]

Should we just be checking feature_config[TYPE] == IMAGE instead?

[jppgks]

It seems like feature type information is not available on the training set metadata