Move information on numerical/non_numerical/encoded_non_numerical from .uns to .var

[eroell]

PR Checklist

• This comment contains a description of changes (with reason)
• Referenced issue is linked
• If you've fixed a bug or added code that should be tested, add tests!
• Documentation in docs is updated

Description of changes
Resolves #620. Uses .var['ehrapy_column_type'] instead of .uns for numeric, non_numeric, and non_numeric_encoded type identification of variables.

Technical details
Instead of .uns['numerical_columns'], .uns['non_numerical_columns'], .uns['non_numerical_encoded_columns'], one column in .var containing values numeric, non_numeric, or non_numeric_encoded is used.

Reading in or transferring csv files, and a wide variety of users interacting with the AnnData object and ehrapy is not affected. However, backwards compatibility is not strictly maintained with this update. E.g. custom modifications to the .uns['numerical_columns'] etc will break with this update.

Additional context
Putting this information to the variable level in .var allows for e.g. slicing, and reduces overhead of keeping .uns in sync with the variables when selecting/moving variables.

Old example of creating a dummy dataset:

import ehrapy as ep
import numpy as np
import pandas as pd
import scanpy as sc

def create_dummy_dataset_numerical_in_obs():
    """
    Create a dummy dataset with numerical and non-numerical variables in obs.
    Also, has numerical variables in .X.
    """
    dummy_obs = {"disease":['A', 'A', 'A', 'A', 'B', 'B', 'B', 'B'],
                "station": ['ICU', 'ICU', 'MICU', 'MICU', 'ICU', 'ICU', 'MICU', 'MICU'],
                "syst_bp_entry": [138, 139, 140, 141, 148, 149, 150, 151],
                "diast_bp_entry": [78, 79, 80, 81, 77, 78, 79, 80]}

    dummy_var = pd.DataFrame({"Unit": ["mg/dl", "kg"],
                              "ehrapy_column_type": ["numerical", "numerical"]})
    dummy_var.index = ['glucose', "weight"]
    dummy_X = np.array([[80, 90, 120, 130, 80, 130, 120, 90],
                        [77, 76, 60, 90, 110, 78, 56, 76]]).T

    adata_dummy = sc.AnnData(X=dummy_X, obs=dummy_obs, var=dummy_var)
    
    adata_dummy.uns['numerical_columns'] = ['glucose', 'weight']
    adata_dummy.uns['non_numerical_columns'] = []
    adata_dummy.uns['encoded_non_numerical_columns'] = []
    
    return adata_dummy

adata_dummy = create_dummy_dataset_numerical_in_obs()

New example of creating a dummy dataset:

import ehrapy as ep
import numpy as np
import pandas as pd
import scanpy as sc

def create_dummy_dataset_numerical_in_obs():
    """
    Create a dummy dataset with numerical and non-numerical variables in obs.
    Also, has numerical variables in .X.
    """
    dummy_obs = {"disease":['A', 'A', 'A', 'A', 'B', 'B', 'B', 'B'],
                "station": ['ICU', 'ICU', 'MICU', 'MICU', 'ICU', 'ICU', 'MICU', 'MICU'],
                "syst_bp_entry": [138, 139, 140, 141, 148, 149, 150, 151],
                "diast_bp_entry": [78, 79, 80, 81, 77, 78, 79, 80]}

    dummy_var = pd.DataFrame({"Unit": ["mg/dl", "kg"]})
    dummy_var.index = ['glucose', "weight"]
    dummy_X = np.array([[80, 90, 120, 130, 80, 130, 120, 90],
                        [77, 76, 60, 90, 110, 78, 56, 76]]).T

    adata_dummy = sc.AnnData(X=dummy_X, obs=dummy_obs, var=dummy_var)
    
    return adata_dummy

adata_dummy = create_dummy_dataset_numerical_in_obs()

[Zethson]

This is great!

We're not using the chatGPT/vscode/something autogenerated types in the docstrings because we only want to use type hints in the function header. Else, it's 2x the maintenance!