better error message when variables are missing from env

[jeromedockes]

closes #2210

import skrub

data_path = skrub.var("data_path", value="data")
learner = data_path.skb.make_learner()
learner.fit({"data_dir": "data"})

skrub._data_ops._data_ops.UninitializedVariable: No value has been provided for 'data_path'
Evaluation of node <Var 'data_path'> failed. See above for full traceback. This node was defined here:
      File "/home/jerome/workspace/scratch/2026/07/scratch.py", line 14, in <module>
        data_path = skrub.var("data_path", value="data")
- Note that preview values passed to initialize skrub variables
  are ignored by default when we pass an explicit 'environment' dictionary,
  for example when calling SkrubLearner.fit({'X': ..., 'y': ...}).
  Please pass a value for 'data_path' in the environment.
  You can also use skrub.var('data_path', value=..., becomes_default=True)
  to always retain the initialization value as a default.
- WARNING: the following keys were passed in the environment but have no corresponding variable in the DataOp:
  ['data_dir']

[GaelVaroquaux]

Thanks a lot!

Is it possible to detect whether a "make_learner" call has been called or not, to have a more precise error message?