Behaviour of precompute transforms with kernel = None default unclear

[matt-graham]

Both the forward and inverse precompute transforms default to a value of None for their kernel argument and the corresponding docstrings indicate this argument is optional:

s2fft/s2fft/precompute_transforms/spherical.py

Lines 14 to 23 in 7cf80bb

	def inverse(
	flm: np.ndarray,
	L: int,
	spin: int = 0,
	kernel: np.ndarray = None,
	sampling: str = "mw",
	reality: bool = False,
	method: str = "jax",
	nside: int = None,
	) -> np.ndarray:

s2fft/s2fft/precompute_transforms/spherical.py

Lines 289 to 298 in 7cf80bb

	def forward(
	f: np.ndarray,
	L: int,
	spin: int = 0,
	kernel: np.ndarray = None,
	sampling: str = "mw",
	reality: bool = False,
	method: str = "jax",
	nside: int = None,
	) -> np.ndarray:

Looking through the underlying implementations though, I cannot spot any logic for handling a kernel is None case with the kernel value passed as is in to a call to einsum.

I think we probably either want to remove the None defaults (and optional labels in the docstrings) or add some logic for automatically computing an appropriate value for kernel if left as the default None?