fix: resolve logging.py issues from code audit

[fwh888]

Summary

This PR fixes issues found during a code audit of the core rdsa-utils modules (see issue #254).

Changes

1. Bitwise AND → Logical AND (logging.py:237)

# Before
if show_records & stop_pipeline:
# After
if show_records and stop_pipeline:

Bitwise & on boolean values works accidentally but is not Pythonic and will behave unexpectedly with non-standard boolean types (e.g., numpy.bool_).

2. try/finally for Spark cache cleanup (log_rows_in_spark_df)

# Before: kwargs['df'].unpersist() was outside the try block
# After: moved into finally block
try:
    df_return = func(*args, **kwargs)
finally:
    if kwargs.get('df') and isinstance(kwargs['df'], SparkDF):
        kwargs['df'].unpersist()

If the decorated function raises an exception, the input DataFrame's cache was never released, causing a persistent Spark memory leak. Added a type guard to avoid AttributeError on non-DataFrame inputs.

Remaining issues from audit

See issue #254 for the full audit report. Other findings (handler validation order, logger init idempotency, MEMORY_ONLY → MEMORY_AND_DISK) are higher-risk changes that require more careful review.

[coatet]

Closed, as we do not accept submissions from users outside the ONS and I suspect this may be an LLM.