add option to terminate PGS constraint solvers based on a least square residual threshold

(for example solverInfo().m_leastSquaresResidualThreshold = 1e-7 and use large m_numSolverIterations
disable sphere-sphere contact cache, it is buggy (some contact point stay in the cache, when sphere penetrates more than total margins)
tweak some gpu demo settings

src/BulletCollision/CollisionDispatch/btSphereSphereCollisionAlgorithm.cpp

@@ -12,6 +12,7 @@ subject to the following restrictions:
 2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
 3. This notice may not be removed or altered from any source distribution.
 */
+#define CLEAR_MANIFOLD 1
 
 #include "btSphereSphereCollisionAlgorithm.h"
 #include "BulletCollision/CollisionDispatch/btCollisionDispatcher.h"