Enable 16bit and 8bit Row/Col Indices in Qnnpack Fully Connected Sparse Op (#85246)
This diff enables using the 16bit and 8bit kernels added in the previous diff.
(This change used to be in D38954842 v11 but was moved into its own diff)
Differential Revision: [D39403164](https://our.internmc.facebook.com/intern/diff/D39403164/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/85246
Approved by: https://github.com/kimishpatel
