Fix edge case behavior of lgamma_r

`lgamma_r` function computes the natural logarithm of the absolute value
of the gamma function. The sign of the gamma function is returned in the
signp argument of `lgamma_r`.

According to the C99 specification, the gamma function should behave as follows:
- If the argument is a negative integer, NaN is returned
- If the argument is ±0, ±∞ is returned

The current implementation in IGC adheres to these specifications but incorrectly
calculates the `signp` argument when the input is ±0.

OpenCL C specification introduces "Additional Requirements Beyond C99":
- lgamma_r(x, signp) returns 0 in signp if x is zero or a negative integer.

Without this change, IGC incorrectly calls the `__spirv_ocl_sign` built-in
function with the result of `gamma(±0)`, which is ±∞. This results in a `signp`
value of ±1, which does not conform to the OpenCL C specification requirement.

Author: aratajew

IGC/BiFModule/Implementation/Math/lgamma_r.cl

@@ -131,7 +131,7 @@ INLINE float SPIRV_OVERLOADABLE SPIRV_OCL_BUILTIN(lgamma_r, _f32_p0i32, )( float
     {
         float g = SPIRV_OCL_BUILTIN(tgamma, _f32, )(x);
         r = __intel_relaxed_isnan(g) ? INFINITY : SPIRV_OCL_BUILTIN(native_log, _f32, )(SPIRV_OCL_BUILTIN(fabs, _f32, )(g));
-        s = SPIRV_OCL_BUILTIN(sign, _f32, )(g);
+        s = (x == 0) ? 0 : SPIRV_OCL_BUILTIN(sign, _f32, )(g);
     }
     signp[0] = s;
     return r;