Bugfix: groupStateHelper uses potentially wrong WellState

[hakonhagland]

When well_state is copied, groupStateHelper should be updated to use that copy. In the master this worked since
groupStateHelper already used the same version of well_state, but the bug was revealed by the refactoring in #6626, see failed jenkins build test for glift1 there. The plan is to rebase #6626 once this is merged and then rerun the tests.

[hakonhagland]

jenkins build this please

[blattms]

What kind of copy (shallow or deep) is this? Is this documented somewhere?

[hakonhagland]

@blattms This is a shallow copy and is intentional and by design. GroupStateHelper only stores pointers, references and scalar values and can safely be copied. An alternative was discussed when we originally refactored WellGroupHelpers, see #6535 (comment). This involved using a const_cast instead of the copy, i.e.
auto well_guard = const_cast<GroupStateHelperType&>(groupStateHelper).pushWellState(well_state_copy) but we found the copy version to be cleaner.