Is your feature request related to a problem? Please describe.

When accessing columns of a table, we may need to check

1. whether a column with the given name exists,
2. whether the column has a compatible type.

Describe the solution you'd like

Polars has a method collect_schema to compute the schema of a DataFrame or LazyFrame. We may, hence, generate code to fetch the schema of a Table from the runner in the background. Once the information is available, we can run further checks without running code.

The schema checks themselves can be described as constraints in the stub language.