Fix Adaround caching bug in torch>=2.5

[quic-kyunggeu]

Problem

The existing intermediate data caching code was very hacky and doesn't work in torch >= 2.5

aimet/TrainingExtensions/torch/src/python/aimet_torch/utils.py

Lines 912 to 914 in de5c4b5

	module_forward_fn = inspect.currentframe().f_back
	assert inspect.getframeinfo(module_forward_fn).function == '_call_impl' # forward function proxy
	kwargs = module_forward_fn.f_locals['kwargs']

Solution

Removed hackiness. Save kwargs directly without assuming anything from the function signature