[Doc] Add GRU new gate calculation difference (#100646)
Summary: Add a note for the calculation difference of GRU new gate `n_t` between PyTorch and original paper.
Fix: #99531
Test Plan: Please see GitHub pipelines.
Differential Revision: D45579790
Pull Request resolved: https://github.com/pytorch/pytorch/pull/100646
Approved by: https://github.com/mikaylagawarecki
