fix: resolve HuggingFace version alias in image URI tag

[giangle-AI]

Summary

• Bug: When constructing image URIs for HuggingFace framework containers, the code used original_version (raw user input like "4.49") instead of resolving it through the version alias mapping ("4.49" → "4.49.0"). This produced incorrect image tags such as transformers4.49 instead of the correct transformers4.49.0, pointing to non-existent container images.
• Fix: Replace _version = original_version with _version = _version_for_config(version, config) on line 247 of image_uris.py so the version alias is properly resolved before being embedded in the image tag.

Before (broken)

>>> image_uris.retrieve(framework='huggingface', region='ap-southeast-1', version='4.49',
...     py_version='py312', image_scope='inference', base_framework_version="pytorch2.6.0",
...     instance_type="ml.c5.xlarge")
'763104351884.dkr.ecr.ap-southeast-1.amazonaws.com/huggingface-pytorch-inference:2.6.0-transformers4.49-cpu-py312-ubuntu22.04'

After (fixed)

>>> image_uris.retrieve(framework='huggingface', region='ap-southeast-1', version='4.49',
...     py_version='py312', image_scope='inference', base_framework_version="pytorch2.6.0",
...     instance_type="ml.c5.xlarge")
'763104351884.dkr.ecr.ap-southeast-1.amazonaws.com/huggingface-pytorch-inference:2.6.0-transformers4.49.0-cpu-py312-ubuntu22.04'

Test plan

• Verified the bug reproduces on master (transformers4.49 in the tag)
• Verified the fix produces the correct output (transformers4.49.0 in the tag)
• Run existing unit tests for image_uris.retrieve with HuggingFace framework
• Verify other HuggingFace version aliases (e.g. 4.37, 4.51) also resolve correctly
• Confirmed all 26 non-aliased full versions (e.g. 4.49.0, 4.51.3) produce correct tags with no regression

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.