fix: prevent integer overflow in LongIsCloseToAssertion by widening to double

Cast `value` to `double` before subtraction and compare against
`(double)_tolerance` so that extreme long values (e.g., MaxValue vs MinValue)
no longer silently wrap around.

Author: thomhurst

TUnit.Assertions/Conditions/NumericIsCloseToAssertion.cs

@@ -209,9 +209,9 @@ protected override Task<AssertionResult> CheckAsync(EvaluationMetadata<long> met
             return Task.FromResult(AssertionResult.Failed($"threw {exception.GetType().Name}"));
         }
 
-        var diff = Math.Abs(value - _expected);
+        var diff = Math.Abs((double)value - _expected);
 
-        if (diff <= _tolerance)
+        if (diff <= (double)_tolerance)
         {
             return AssertionResult._passedTask;
         }