fix(pipeline): default missing feature configs

[RerankerGuo]

Motivation

Fixes #5317.

MineruPipelineModel assumes both formula_config and table_config are
dictionaries. Direct integrations that omit either optional config, or pass
it as None, fail before model initialization:

AttributeError: 'NoneType' object has no attribute 'get'

The standard CLI supplies both dictionaries, but the model class accepts
keyword arguments directly and should preserve its documented default-enabled
behavior when those optional dictionaries are absent.

Modification

• Normalize a missing or None formula_config to an empty dictionary.
• Normalize a missing or None table_config to an empty dictionary.
• Add regression coverage for omitted configs, explicit None, and explicit
  {"enable": False} values.
• Mock atom model creation in the tests, so no weights are downloaded and no
  GPU is required.

BC-breaking (Optional)

No. Missing configs now use the existing default of enable=True. Explicit
enable=False values remain unchanged.

Verification

The same regression test was run against the original and patched code.

Before:

2 failed, 1 passed
AttributeError: 'NoneType' object has no attribute 'get'

After:

3 passed

Commands:

PYTHONPATH=. uv run --python 3.12 --with pytest --with torch \
  --with torchvision --with loguru --with numpy --with opencv-python \
  --with transformers --with modelscope --with pypdfium2 --with pypdf \
  --with pdftext --with pillow --with shapely --with pyclipper \
  --with onnxruntime --with fast-langdetect --with six \
  --with beautifulsoup4 --no-project pytest -q -o addopts='' \
  tests/unittest/test_pipeline_model_config.py

uvx ruff check --ignore ANN mineru/backend/pipeline/model_init.py
uvx ruff check tests/unittest/test_pipeline_model_config.py

No model or GPU end-to-end run is needed for this constructor-level change;
the regression test replaces all atom model factories and exercises the
configuration boundary directly.

Checklist

Before PR:

• Pre-commit or other linting tools are used to fix the potential lint issues.
• Bug fixes are fully covered by unit tests, the case that causes the bug should be added in the unit tests.
• The modification is covered by complete unit tests.
• Documentation impact was reviewed; no public API documentation change is needed.

After PR:

• No downstream API or model behavior changes for explicit configurations.
• CLA has already been signed by this contributor.