default_ml_model() signature stale: TypeError 'task_cls' breaks every task that resolves ml_models

[artlbv]

default_ml_model() signature is stale: any task resolving ml_models dies with TypeError: unexpected keyword argument 'task_cls'

Symptom

Any task that resolves the ml_models default -- cf.PlotVariables1D, hbw.CustomCreateYieldTable,
and every other MLModelsMixin task -- fails during parameter resolution, before any work starts:

File "columnflow/tasks/framework/mixins.py", line 1643, in resolve_param_values_pre_init
    params["ml_models"] = cls.resolve_config_default_and_groups(
File "columnflow/tasks/framework/base.py", line 791, in resolve_config_default_and_groups
File "columnflow/tasks/framework/base.py", line 681, in resolve_config_default
    _param = _param(task_cls=cls, container=_container, task_params=task_params)
TypeError: default_ml_model() got an unexpected keyword argument 'task_cls'

Observed with columnflow v0.3.1-49-gd20db156.

Cause

columnflow calls the config's default_ml_model callable with keyword arguments:

# columnflow/tasks/framework/base.py:681
_param = _param(task_cls=cls, container=_container, task_params=task_params)

but hbw declares the first parameter as cls:

# hbw/config/sl_defaults_and_groups.py:39   (and hbw/config/defaults_and_groups.py:48)
def default_ml_model(cls, container, task_params):

Both are registered as config_inst.x.default_ml_model
(sl_defaults_and_groups.py:127, defaults_and_groups.py:123), so both are affected.

Suggested fix

Rename the first parameter to task_cls in both definitions, and accept **kwargs so a future
signature change upstream does not break it again:

def default_ml_model(task_cls, container, task_params, **kwargs):
    cls = task_cls   # keep the existing body unchanged

The body uses cls only via hasattr(cls, "ml_model"), so the rename is local.

Workaround

Pass --ml-models "" explicitly, which skips default resolution entirely.

Context

Hit while running the TOPO trigger sensitivity study (sl1_topo selector, l24v15); not specific
to that setup -- it reproduces on any plotting task that does not pin --ml-models.