Move prim::tolist math.log and aten::cpu to lite interpreter for translation model (#45482)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/45482
Working on some models that need these ops on lite interpreter.
Test Plan: locally build and load/run the TS model without problem.
Reviewed By: iseeyuan
Differential Revision: D23906581
fbshipit-source-id: 01b9de2af2046296165892b837bc14a7e5d59b4e
