fix(util): avoid passing None to load_yaml on failed jinja render

[sundeep8967]

Problem

When a templated YAML config cannot be rendered (for example, when referencing a missing nested attribute like {{ ds.meta_data.placement.region }} on a platform where it is not defined), render_jinja_payload_from_file() logs a warning and returns None.

In util.read_conf(), config_file was being unconditionally overwritten with this None return value:

config_file = render_jinja_payload_from_file(...)

Subsequently, load_yaml(config_file) was called with config_file=None, resulting in decode_binary(None) raising:

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

and causing the init-local / init boot stage to fail.

Solution

1. In util.read_conf(), capture the result of render_jinja_payload_from_file() in a temporary variable and only update config_file if the rendered output is not None.
2. Added unit test test_read_conf_with_failed_jinja_render_undefined_attribute in tests/unittests/test_util.py asserting that when jinja rendering fails due to an undefined attribute, the original configuration is preserved and loaded without raising an AttributeError.

Fixes #7007

Testing

• Verified test failure prior to the fix (AttributeError on decode_binary(None) reproduced).
• Ran pytest tests/unittests/test_util.py — all 318 tests passed.