Using real_t for toMatrix does not work with types that are not real_t
Within the file `hog/operator_generation/kernel_types.py` a generic field called mat is created and its `dtype` is set to `real_t`. This only works if everything is `real_t`, when you want to use other types, e.g. for mixed precision, compile errors occur, since `real_t` (usually float64) cannot be converted to a 'smaller' type.
> cannot convert '`walberla::real_t*`' {aka '`double*`'} to '`const float*`'
The question is, what is the desired behavior. I understand that the signature is given as `real_t` and maybe it is mathematically advised to even use float64, but this way it is not possible to experiment with different precisions.
---
Code part of interest (kernel_types.py line 268 to 276) [link-to-code](https://i10git.cs.fau.de/hyteg/hog/-/blob/main/hog/operator_generation/kernel_types.py?ref_type=heads#L268):
```Python
# NOTE: The type is 'hyteg_type().pystencils_type', i.e. `real_t`
# (not `self._type_descriptor.pystencils_type`)
# because this function is implemented manually in HyTeG with
# this signature. Passing `np.float64` is not ideal (if `real_t !=
# double`) but it makes sure that casts are inserted if necessary
# (though some might be superfluous).
mat = create_generic_fields( ["mat"], dtype=hyteg_type(HOGPrecision.REAL_T).pystencils_type )[0]
```
issue