adjustment should be allowed when desired is in the scope of min_size/max_size when doing cluster resizing

There is a bug in current implementation of truncate_desired(): the cluster's min_size and max_size should only be used when min_size/max_size input parameter is None. Otherwise, the truncation will always be done using cluster's current size scope.

def truncate_desired(cluster, desired, min_size, max_size):
    '''Do truncation of desired capacity for non-strict cases.'''

    if min_size is not None and desired < min_size:
        desired = min_size
        LOG.debug(_("Truncating shrinkage to specified min_size (%s).")
                  % desired)

    if desired < cluster.min_size:
        desired = cluster.min_size
        LOG.debug(_("Truncating shrinkage to cluster's min_size (%s).")
                  % desired)

    if (max_size is not None and max_size > 0 and desired > max_size):
        desired = max_size
        LOG.debug(_("Truncating growth to specified max_size (%s).")
                  % desired)

    if desired > cluster.max_size and cluster.max_size > 0:
        desired = cluster.max_size
        LOG.debug(_("Truncating growth to cluster's max_size (%s).")
                  % desired)

    return desired