ensure correct behaviour of filters when set manually

[JonasHell]

When set manually via filter_config, there have been some issues with filters of type structureTensorEigenvalues since they require the additional argument outerScale.

raw = np.random.rand(32, 256, 256)
filters_and_sigmas = _get_filters(3, [('hessianOfGaussianEigenvalues', (3., 1.25*3, 1.25*3))])
features = _apply_filters(raw, filters_and_sigmas)
print(features.shape)

--> works fine

filters_and_sigmas = _get_filters(3, [('structureTensorEigenvalues', (3., 1.25*3, 1.25*3))])
features = _apply_filters(raw, filters_and_sigmas)
print(features.shape)

--> throws the following error

---------------------------------------------------------------------------
File ~/torch-em/torch_em/shallow2deep/prepare_shallow2deep.py:222, in _apply_filters(raw, filters_and_sigmas)
    217     func = getattr(filter_impl, filter_) if isinstance(filter_, str) else filter_
--> 218 response = func(raw, sigma)
    219 if response.ndim > raw.ndim:
    220     for c in range(response.shape[-1]):

TypeError: structureTensorEigenvalues() missing 1 required positional argument: 'outerScale'

--> fixed now