[BUG] Rare case with custom adapter - standardization modifies the original data

[stefanradev93]

Describe the bug
In the rare case of a power user using no adapter (i.e., providing an identity function as an adapter) AND any kind of standardization in the workflow, the original data in conditions during sampling (and for that matter, in the training set) will be surreptitiously changed (i.e., standardized) without the user ever noticing!

Expected behavior
Approximator objects should have no side effects on conditions or datasets regardless of the adapter's behavior (copy or not). A shallow copy should always be used.