Ridiculously large delta for SwarmSolverTest::testUnconstrained for now

At least the Jenkins lo_ubsan build sporadically keeps failing
CppunitTest_sccomp_swarmsolvertest due to rather widely off values here
(3.00064356398118 in <https://ci.libreoffice.org/job/lo_ubsan/736/console> and
2.8639022684852 in <https://ci.libreoffice.org/job/lo_ubsan/738/console>), so
suggestion by Tomaž Vajngerl was: "Let's adapt the delta for now. Generally
anything close to 3 should be acceptable as the algorithm greatly depends on
random values."

Change-Id: Id8f8657e521381decb1c719a1cffdf959ae74ef4
Reviewed-on: https://gerrit.libreoffice.org/45231Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Tomaž Vajngerl <quikee@gmail.com>

sccomp/qa/unit/SwarmSolverTest.cxx

@@ -103,7 +103,7 @@ void SwarmSolverTest::testUnconstrained()
     uno::Sequence<double> aSolution = xSolver->getSolution();
 
     CPPUNIT_ASSERT_EQUAL(aSolution.getLength(), aVariables.getLength());
-    CPPUNIT_ASSERT_DOUBLES_EQUAL(3.0, aSolution[0], 1E-5);
+    CPPUNIT_ASSERT_DOUBLES_EQUAL(3.0, aSolution[0], .2);
 }
 
 void SwarmSolverTest::testVariableBounded()