Use closer guess for MacOSX failing test.

This was failing on MacOSC in these versions:

Python 3.8 + IPOPT 3.13 & 3.14
Python 3.9 + IPOPT 3.13
Python 3.10 + IPOPT 3.13 & 314

with this error:

________________ TestSLSQP.test_minimize_unbounded_approximated ________________

self = <cyipopt.tests.unit.test_scipy_ipopt_from_scipy.TestSLSQP object at 0x115154a30>

    def test_minimize_unbounded_approximated(self):
        # Minimize, method=None: unbounded, approximated jacobian.
        jacs = [None, False]
        for jac in jacs:
            res = minimize(self.fun, [-1.0, 1.0], args=(-1.0, ),
                           jac=jac, method=None,
                           options=self.opts)
>           assert_(res['success'], res['message'])
E           AssertionError: b'Algorithm stopped at a point that was converged, not to "desired" tolerances, but to "acceptable" tolerances (see the acceptable-... options).'

cyipopt/tests/unit/test_scipy_ipopt_from_scipy.py:158: AssertionError
=========================== short test summary info ============================
FAILED cyipopt/tests/unit/test_scipy_ipopt_from_scipy.py::TestSLSQP::test_minimize_unbounded_approximated - AssertionError: b'Algorithm stopped at a point that was converged, not to "desired" tolerances, but to "acceptable" tolerances (see the acceptable-... options).'
====== 1 failed, 90 passed, 12 skipped, 1 xfailed, 43 warnings in 11.88s =======
Error: Process completed with exit code 1.

I changed the guess for this test to possibly give convergence.

cyipopt/tests/unit/test_scipy_ipopt_from_scipy.py

@@ -170,7 +170,7 @@ def test_minimize_bounded_approximated(self):
         jacs = [None, False]
         for jac in jacs:
             with np.errstate(invalid='ignore'):
-                res = minimize(self.fun, [-1.0, 1.0], args=(-1.0, ),
+                res = minimize(self.fun, [2.0, 1.0], args=(-1.0, ),
                                jac=jac,
                                bounds=((2.5, None), (None, 0.5)),
                                method=None, options=self.opts)