Fix unwanted behavior of CLP when we expect to do dual simplex

CLP is hinted to go for dual simplex but it may still go for primal, giving wrong
results when in strong branching iteration limit is hit and primal has been done.
Moreover, when doing dual simplex, dual solutions may be inaccure at iteration limit.
To fix that, some options are set through Osi to make CLP behave as we expect.

src/LP/lp_solver.c

@@ -2339,11 +2339,22 @@ void open_lp_solver(LPdata *lp_data)
    lp_data->si->messageHandler()->setLogLevel(0);
 #ifdef __OSI_CLP__
    lp_data->si->setupForRepeatedUse();
-   //lp_data->si->setupForRepeatedUse(3,0);
-   //lp_data->si->getModelPtr()->setFactorizationFrequency(200);
-   //lp_data->si->getModelPtr()->setSparseFactorization(true);
-   //lp_data->si->getModelPtr()->setSpecialOptions(524288);
-   //lp_data->si->getModelPtr()->setSpecialOptions(4);   
+   // feb223
+   int options = lp_data->si->getModelPtr()->specialOptions();
+   int moreOptions = lp_data->si->getModelPtr()->moreSpecialOptions();
+
+   options |= 32           // CLP creates ray while doing dual simplex
+            // + 2048         // CLP does not cruch the problem
+                              // maybe not needed if we recompute dj's on fixed vars ???
+            + 0x08000000;  // CLP computes accurate duals on LP_D_ITLIM
+   moreOptions |= 8192;    // CLP uses dual simplex if it has suggested to do so
+   lp_data->si->getModelPtr()->setSpecialOptions(options);
+   lp_data->si->getModelPtr()->setMoreSpecialOptions(moreOptions);
+   // lp_data->si->setupForRepeatedUse(3,0);
+   // lp_data->si->getModelPtr()->setFactorizationFrequency(200);
+   // lp_data->si->getModelPtr()->setSparseFactorization(true);
+   // lp_data->si->getModelPtr()->setSpecialOptions(524288);
+   // lp_data->si->getModelPtr()->setSpecialOptions(4);
    lp_data->si->getModelPtr()->setPerturbation(50);
    //set cleanup param if unscaled primal is infeasible
    lp_data->si->setCleanupScaling(1);