[ColumbusLabs]

Fixes #36

SpectralPooling1D and SpectralPooling2D could not be used with TensorFlow's Keras:

• call() read x._keras_shape, which only standalone Keras 1 set, so both layers raised AttributeError when called.
• SpectralPooling1D.__init__ took topf as a named argument with an unused default and then looked for it in **kwargs, so SpectralPooling1D(topf=...) always raised "Must provide either topf= or gamma= !".
• Neither layer passed its kwargs to Layer.__init__, so name= and the other standard layer arguments were dropped.

Changes (complexnn/pool.py only)

• Both layers read the static shape with tuple(x.shape) via a small _static_shape helper, which raises a ValueError if a spatial dimension is None (the masks are built from the static size).
• Both __init__ methods take topf=None and gamma=None explicitly and forward the remaining kwargs to super().__init__. Calling with neither still raises the same RuntimeError.
• The fft module is untouched.

Tests

• New tests/test_pool.py: builds and trains a one-step model with each layer using topf= and gamma=, checks the output mask, checks name= is forwarded, and checks the ValueError for an unknown spatial dimension.
• TestNoKerasShapeWrites in tests/test_public_api.py is now TestNoKerasShapeUse and also catches reads of _keras_shape, not just writes.

pytest tests/test_pool.py tests/test_public_api.py:

Configuration	main (new tests applied)	this branch
Python 3.13, TensorFlow 2.21.0 (Keras 3)	8 failed, 3 passed	11 passed
Python 3.11, TensorFlow 2.15.1 + tf-keras, TF_USE_LEGACY_KERAS=1	8 failed, 3 passed	11 passed

ruff check and ruff format --check pass on the changed files.