E3nn-jax

Changed
- **[BREAKING]** Renamed `e3nn.util` in `e3nn.utils`

Added
- `Irreps.set_mul(int)` to set the multiplicity of all irreps
- `Irreps.filter(lmax=int)` to filter out irreps with `l > lmax`
- `IrrepsArray.filter(lmax=int)` to filter out irreps with `l > lmax`
- `IrrepsArray.__radd__` and `IrrepsArray.__rsub__` to support `scalar + IrrepsArray` and `scalar - IrrepsArray`
- `0 + IrrepsArray` and `0 - IrrepsArray` are now always accepted as special cases.
- Support for `IrrepsArray / array`
- Add `utils` as a submodule

Fixed
- `e3nn.scatter` operation handle indices with `ndim > 1`

Added
- `e3nn.cross` for completeness

Changed
- Optimize `e3nn.reduced_symmetric_tensor_product_basis`, especially for the `keep_ir` argument

Added
- `LinearSHTP` module implementing the optimized linear mixing of inputs tensor product with spherical harmonics
- `D_from_axis_angle`
- `to_s2grid`: `quadrature="gausslegendre"` by default
- `soft_odd` activation function for odd scalars
- more support of arrays implicitely converted into `IrrepsArray` as scalars (i.e. added few `IrrepsArray.as_irreps_array`)

Changed
- `scalar_activation` simpler to use with default activation functions (a bit like gate)

Changed
- `e3nn.normalize_function` now uses a deterministic (not pseudorandom) algorithm to compute the normalization factor.

Added
- `normalize_act` option to `e3nn.scalar_activation` and `e3nn.gate`. We can now turn the normalization off if we want to.
- `e3nn.norm_activation` as a new activation function.

Fixed
- Fix `NaN` in the gradients of `e3nn.xyz_to_angles`. The gradients are now `0` when the input is on the poles.

Added
- `e3nn.dot`: compute the dot product between two `IrrepsArray`
- `per_irrep` argument to `e3nn.norm`: compute the norm of each irrep independently if `per_irrep=True`
- `e3nn.tensor_product_with_spherical_harmonics` from https://arxiv.org/pdf/2302.03655.pdf

Changed
- `__repr__(Irreps())` has been changed from `""` to `"Irreps()"`

Fixed
- spherical harmonics edge case when `output_irreps=Irreps()`