Fix failing tests: decimate in chunks and skops

[alejoe91]

Ported here some test fixes from #4713 and #4703

[chrishalcrow]

From the scikit-learn warning message:

- sklearn.tree._tree.Tree: sklearn.tree._tree.Tree (the shared node storage for DecisionTree*, RandomForest*, ExtraTrees*, and GradientBoosting* models) stores raw node indices (left_child, right_child, feature) that scikit-learn indexes into without bounds checking. A malicious file can set these to out-of-range values: skops loads the object successfully, but calling .predict() on it can then crash the process (segfault) or read out-of-bounds memory. If you created the file yourself or otherwise fully trust its source, you can load it with trusted=["sklearn.tree._tree.Tree"].
    Only add the specific types you have reviewed and trust to the `trusted` argument; avoid passing everything reported by get_untrusted_types() just to make a file load.

I think it's probably ok, because we force the user to set trust_model = True so it's their responsibility. But it's a bit awkward.

[alejoe91]

that's the same strategy we had before, I dodn't know there was a function for it so I was parsing the exception...

[chrishalcrow]

The idea of the exception was to force users to think for a second before blindly trusting.