Fix overflow approximating fractions of large negative constants

FractionApprox rejects inputs above 0x7FFFFFFF because they cannot be stored
as a fraction, but the check was one-sided while the sign is stripped right
after it. A large negative value therefore reached the continued-fraction loop
and overflowed the terms it accumulates. ICSharpCode.Decompiler is built with
CheckForOverflowUnderflow, so that aborted decompilation of the whole member
instead of wrapping.

Found by fuzzing nuget.org; reproduces on MathNet.Numerics, whose constants
reach the approximation through their ratio to PI.

Assisted-by: Claude:claude-opus-5[1m]:Claude Code

ICSharpCode.Decompiler.Tests/TestCases/Pretty/WellKnownConstants.cs

@@ -202,5 +202,10 @@ namespace ICSharpCode.Decompiler.Tests.TestCases.Pretty
 		public const double Double_Negated_E = -Math.E;
 		public const double Double_Negated_BeforeE = -2.7182818284590446;
 		public const double Double_Negated_AfterE = -2.7182818284590455;
+
+		// Values of either sign whose ratio to PI or E is outside the range expressible as
+		// a fraction must be left alone rather than run through the approximation.
+		public const double Double_TooLargeForFraction = 3085105840164255.5;
+		public const double Double_Negated_TooLargeForFraction = -3085105840164255.5;
 	}
 }

ICSharpCode.Decompiler/CSharp/Syntax/TypeSystemAstBuilder.cs

@@ -1724,7 +1724,10 @@ namespace ICSharpCode.Decompiler.CSharp.Syntax
 		// we just keep the last partial product of these matrices.
 		static (long Num, long Den) FractionApprox(double value, int maxDenominator)
 		{
-			if (value > 0x7FFFFFFF)
+			// The range check has to be on the magnitude: the sign is stripped below, so a
+			// large negative value would otherwise reach the continued-fraction loop and
+			// overflow the terms it accumulates.
+			if (Math.Abs(value) > 0x7FFFFFFF)
 				return (0, 0);
 
 			double startValue = value;