Handle DAC conversion when using weight_norm with newer PyTorch versions

[edwko]

What does this PR do?

nn.utils.weight_norm has been moved to nn.utils.parametrizations.weight_norm. As a result, parametrized layers no longer expose weight_g or weight_v, but only weight. This causes an AttributeError when DAC conversion attempts to access weights:

AttributeError: 'ParametrizedConv1d' object has no attribute 'weight_g'. Did you mean: 'weight'?

This PR moves apply_weight_norm directly into the script, explicitly setting to use weight_norm = nn.utils.weight_norm to ensure compatibility.

[xenova]

cc @kamilakesbi, who made the original conversions (do they still work with the existing conversion scripts?)