Fix dequantize_per_channel for single dimension input tensor (#3867)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/3867

This diff has two fixes:
- When the input dimension is 1, we need to handle that separately
- Fixing logic in how dim_list is generated

Reviewed By: larryliu0820

Differential Revision: D58221156

fbshipit-source-id: aee6b66952271e2724c27cb4efd04bc4f54434b1