[FEATURE] Columnar data creation helper function

[nitbharambe]

A initialize_column(data_type, component, shape, attributes, empty=False) helper function which gives out columnar data ie. dict of arrays.

Motivation:

We have initialize_array helper function to make users not bother about the dtype. A same one for columnar data.

If the user has to create a columnar data now, they would have to do :
{k: v for k,v in initialize_array(dataset_type, component_type, shape).items()}
or
{k: np.empty(shape=..., dtype=power_grid_meta_data[dataset_type][component_type][k]) for k in attributes}
or some similar combination

Also in first way, user creates a huge array; subset of which is then converted to dictionary.

Options:

initialize_column(data_type, component, shape, attribute, empty=False) -> np.ndarray

initialize_columnar_data(data_type, component, shape, attributes, empty=False) -> dict[str, np.ndarray]

[mgovers]

decision: use initialize_array(data_type, component, shape, empty=False, *, attributes=None) with additional kwarg

[petersalemink95]

decision: use initialize_array(data_type, component, shape, empty=False, *, attributes=None) with additional kwarg

To be clear: attributes=None should refer to the row based initialization, that already is present. For columnar people should specify a "filter" in attributes=, similar as in the data filter

[nitbharambe]

Update: This helper function would make assignment operations ambiguous for users. Hence it is now decided to have attribute_dtype(dataset_type, component_type, attribute) and attribtue_empty functions.

[Jerry-Jinfeng-Guo]

@nitbharambe Is this still open?