Fix #1410, Apply fabs to full relative error in UtAssert_DoubleCmpRel

UtAssert_DoubleCmpRel computed fabs((x) - (y)) / (x), applying fabs only
to the numerator. When x is negative the quotient becomes negative, so
the comparison (negative <= Ratio) is always true and the assertion
passes regardless of how far apart x and y are (false positive).

Wrap the entire relative-error expression in fabs so the comparison uses
the magnitude of the relative difference: fabs(((x) - (y)) / (x)).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: sg20180546

ut_assert/inc/utassert.h

@@ -198,7 +198,7 @@ typedef struct
  * \brief Compares two floating point numbers and determines if they are equal within a specified relative tolerance.
  */
 #define UtAssert_DoubleCmpRel(x, y, Ratio, ...) \
-    UtAssertEx((fabs((x) - (y)) / (x) <= (Ratio)), UtAssert_GetContext(), __FILE__, __LINE__, __VA_ARGS__)
+    UtAssertEx((fabs(((x) - (y)) / (x)) <= (Ratio)), UtAssert_GetContext(), __FILE__, __LINE__, __VA_ARGS__)
 
 /**
  * \brief  Compares two strings and determines if they are equal.