Fix `EfficientViT_CLS` workload projection

[ramamalladiTT]

Issue:

failed with Cannot squeeze dimension 2 with size 8 (must be 1)
...
"ttsim/ops/desc/tensor.py", line 157, in squeeze_sinf
    raise ValueError(f"Cannot squeeze dimension {idx} with size {dataT.shape[idx]} (must be 1)")
ValueError: Cannot squeeze dimension 2 with size 8 (must be 1)

How to reproduce:

$ python polaris.py -w config/ip_workloads.yaml -a config/all_archs.yaml -m config/wl2archmapping.yaml --filterwlg ttsim --filterwl EfficientViT_CLS -o ODDIR_ttsim -s SIMPLE_RUN --outputformat json

[ssapreTT]

This check was introduced recently, added since squeeze removes specified dimensions of input of size 1, when dimensions are explicitly specified.

This failing run seems to explicitly remove dimension 2, with a non-1 value of 8. Is there some difference in the usage of squeeze operator?