1. It's no problem with relative precision 1e-3 when testing several times in my local machine.
  2. But the testing failed with 1e-2 in the TeamCity, and only one value's relative precision is over 1e-2. So tune it to 4e-2

Fit Google C++ Style

* add_op could take a unique_ptr or a const reference. If unique_ptr is
  taken, the NetOp will take care of that operator's life cycle. If a
  const reference is taken, that op will be Cloned.

* Also simplify pybind implementation by using OperatorBase as holder
  type.