move embeddinglookup to the internal

Merge pull request !2424 from xutianchun/ctcloss

Merge pull request !2369 from baihuawei/reduce

- It provides a unified style to express graph and kernel for user.
- It provides a unified IR to represent graph and kernel for developer.
- It breaks the boundary between graph and kernel.
- It provides more opportunities to do compile optimization.

Merge pull request !2338 from chenweifeng/nezha

Merge pull request !2379 from zyli2020/change_ftrl_op_st

Merge pull request !2282 from chenzhongming/master

Merge pull request !2281 from zyli2020/add_sigmoid

Merge pull request !2292 from limingqi107/master

Merge pull request !2194 from 王东旭/master

Merge pull request !2251 from YuJianfeng/proximal

Merge pull request !2150 from chenweifeng/tanh-fp16

Merge pull request !2229 from zyli2020/add_ftrl_op

Merge pull request !2226 from zyli2020/add_adam_op

Merge pull request !2235 from zyli2020/add_sigmoid_cross_entropy_with_logits_grad

Merge pull request !2234 from zyli2020/add_sigmoid_cross_entropy_with_logits

Merge pull request !2210 from limingqi107/max_device_memory_optimize

Merge pull request !2051 from chenjianping/host_reduce

Merge pull request !2093 from VectorSL/argmaxwithvalue

Merge pull request !2021 from VectorSL/gpu-add-akg-kernel

Merge pull request !1888 from YuJianfeng/adam

Merge pull request !1905 from baihuawei/cpulstm