MixtureFugacityTP::solveCubic finds wrong root on mingw

[ischoegl]

Problem description

Within the test suite, TEST_F(cubicSolver_Test, solve_cubic) fails on mingw, as discussed on #1144. For mingw, the check for

EXPECT_NEAR(nSolnValues, -2, 1.e-6);

fails, where +2 is returned instead. At the moment, the test is disabled (ensuring that the newly introduced mingw GH runner succeeds).

Additional context

The PengRobinson and RedlichKwong models both use this value and do not treat the positive and negative return values equally.

Issue goes back to #1047

An associated test is disabled for mingw:

cantera/test/thermo/cubicSolver_Test.cpp

Lines 26 to 34 in e5122f0

	#ifdef __MINGW32__
	TEST_F(cubicSolver_Test, solve_cubic_DISABLED)
	{
	// the following test fails on mingw: EXPECT_NEAR(nSolnValues, -2, 1.e-6);
	// where the positive root is found instead
	}
	#else
	TEST_F(cubicSolver_Test, solve_cubic)
	{