Rules using dynamic column selection can yield unexpected results

[AndreasAlbertQC]

Inspired by #331 (comment)

Here's an example with a rule that relies on dynamically selecting all boolean columns.
If there are no group rules, this works as expected.
Adding a group rule can make validation fail unexpectedly. I have not debugged this in detail, but I thikn the reason is that we chain with_columns here in a way that makes the rule expression "see" the result of the group rule evaluation.

MWE:

import polars as pl

import dataframely as dy


class MySchema(dy.Schema):
    x = dy.Bool()

    @dy.rule()
    def myrule(cls):
        return ~pl.any_horizontal(pl.col(pl.Boolean))


class MySchemaWithGroupRule(MySchema):
    @dy.rule(group_by="x")
    def my_group_rule(cls):
        return pl.lit(True)


df = pl.DataFrame({"x": [False]})

# Succeeds
MySchema.validate(df)

# Fails
MySchemaWithGroupRule.validate(df)

Result:

  File "/Users/x/repos/dataframely/test.py", line 26, in <module>
    MySchemaWithGroupRule.validate(df)
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^
  File "/Users/x/repos/dataframely/dataframely/schema.py", line 573, in validate
    raise ValidationError(
        format_rule_failures(list(failure.counts().items()))
    )
dataframely.exc.ValidationError: 1 rules failed validation:
 - 'myrule' failed for 1 rows

Expectation: validation should succeed in both cases.