Fix concat compile warning

[chengduoZH]

Fix concat compile warning

../paddle/fluid/operators/math/concat.cc: In instantiation of ‘void paddle::operators::math::ConcatGradFunctor<paddle::platform::CPUDeviceContext, T>::operator()(const paddle::platform::CPUDeviceContext&, const paddle::framework::Tensor&, const std::vector<const paddle::framework::Tensor*>&, int, std::vector<paddle::framework::Tensor*>*) [with T = int]’:
../paddle/fluid/operators/math/concat.cc:118:16:   required from here
../paddle/fluid/operators/math/concat.cc:96:26: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
     for (size_t k = 0; k < input_rows; ++k) {
                          ^
../paddle/fluid/operators/math/concat.cc:99:25: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
       for (int j = 0; j < num; ++j) {
                         ^
../paddle/fluid/operators/math/concat.cc: In instantiation of ‘void paddle::operators::math::ConcatGradFunctor<paddle::platform::CPUDeviceContext, T>::operator()(const paddle::platform::CPUDeviceContext&, const paddle::framework::Tensor&, const std::vector<const paddle::framework::Tensor*>&, int, std::vector<paddle::framework::Tensor*>*) [with T = long int]’:
../paddle/fluid/operators/math/concat.cc:119:16:   required from here
../paddle/fluid/operators/math/concat.cc:96:26: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
     for (size_t k = 0; k < input_rows; ++k) {
                          ^
../paddle/fluid/operators/math/concat.cc:99:25: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
       for (int j = 0; j < num; ++j) {
                         ^
../paddle/fluid/operators/math/concat.cc: In instantiation of ‘void paddle::operators::math::ConcatGradFunctor<paddle::platform::CPUDeviceContext, T>::operator()(const paddle::platform::CPUDeviceContext&, const paddle::framework::Tensor&, const std::vector<const paddle::framework::Tensor*>&, int, std::vector<paddle::framework::Tensor*>*) [with T = float]’:
../paddle/fluid/operators/math/concat.cc:120:16:   required from here
../paddle/fluid/operators/math/concat.cc:96:26: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
     for (size_t k = 0; k < input_rows; ++k) {
                          ^
../paddle/fluid/operators/math/concat.cc:99:25: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
       for (int j = 0; j < num; ++j) {
                         ^
../paddle/fluid/operators/math/concat.cc: In instantiation of ‘void paddle::operators::math::ConcatGradFunctor<paddle::platform::CPUDeviceContext, T>::operator()(const paddle::platform::CPUDeviceContext&, const paddle::framework::Tensor&, const std::vector<const paddle::framework::Tensor*>&, int, std::vector<paddle::framework::Tensor*>*) [with T = double]’:
../paddle/fluid/operators/math/concat.cc:121:16:   required from here
../paddle/fluid/operators/math/concat.cc:96:26: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
     for (size_t k = 0; k < input_rows; ++k) {
                          ^
../paddle/fluid/operators/math/concat.cc:99:25: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
       for (int j = 0; j < num; ++j) {

[luotao1]

LGTM