Improve error handling of classifiers and regressors

[lars-reimann]

Is your feature request related to a problem?

In some cases, error handling is not ideal (see solution for details) in the safeds.ml package.

Desired solution

• Validate inputs of a function as the first step
• Raise an appropriate exception with an appropriate message if issues are found
• Document in the docstring that the exception is raised and under which condition
• Test that the correct exception with the correct message is raised

Possible alternatives (optional)

No response

Screenshots (optional)

No response

Additional Context (optional)

• Example: If non-numerical columns are in a table, an appropriate message should be shown. Users should also be guided towards the OneHotEncoder or LabelEncoder and we should describe when to use which.