[TorchTidy] Add pattern to detect if single-tensor implementation optimizer is used (#81733)
Summary: For Adam, SGD, AdamW, enabling multi-tensor implementation is faster. So for these optimizers, we detect if single-tensor implementation is used.
Test Plan:
Added test in test_profiler.py
Differential Revision: [D38082641](https://our.internmc.facebook.com/intern/diff/D38082641)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/81733
Approved by: https://github.com/robieta
