Fix and enable failed gradient signal tests on Linux

* Fix failure reason for PerlinGradientComponent_GoldenTest: non-standard behavior of std::shuffle producing different results across platforms
* Fix failure reason for GradientImageAssetConversionU8SingleScale: Undefined behavior when float casts to a uint8 has overflow

Signed-off-by: Steve Pham <82231385+spham-amzn@users.noreply.github.com>

Gems/GradientSignal/Code/Tests/ImageAssetTests.cpp

@@ -111,7 +111,10 @@ namespace
         constexpr auto numChannels = 1;
         constexpr auto bytesPerPixel = numChannels * sizeof(AZ::u8);
         constexpr auto outputSize = imageDimensions * imageDimensions;
-        constexpr auto scaling = 25;
+        // Adjust the test scale so that none of the input data overflows the cast to AZ::u8. 
+        // The overflow behavior when casting from float to uint8 is undefined so we want to 
+        // avoid that consistently across all platforms
+        constexpr auto scaling = 255.0f / aznumeric_cast<float>(imageDimensions * imageDimensions);
 
         auto inputData = Detail::GenerateInput<AZ::u8, imageDimensions, numChannels>(scaling);