Fix: decode JSON type before to_list or to_dict is called

[ItsTania]

Motivation:
There is a change in the approach to decoding JSON types, added in d560b58. Since 4.7.0, to_list() returns raw JSON strings for columns stored as Json(), while direct access via getitem returns dicts.

In versions prior to 4.7.0, to_list() returned dicts.

Added context:

• The issue outlining the breaking change is datasets.to_list: Behaviour change in version 4.7.0 with the addition of JSON type #8131
• On it, the maintainer suggested taking a similar approach to fix: decode Json() values before calling DataFrame.to_json() (#8116) #8122

This PR contains:

• An update to the to_list and to_dict functions for the arrow dataset inspired by a (and using the utils developed for) similar fix to another problem caused by the addition of the JSON type.
• Unit tests specifying the intended outcome (which was how dictionaries were processed or handled prior to v4.7.0).

Additional questions/considerations:

• This PR restore behaviour prior to the breaking change introduced by the addition of the JSON datatype in v4.7.0. However, does mean we now decode one data type (JSON) and not others. We could decide to decode elements of the list in the same way _get_item does but this would be a big change (i.e memory concerns with images and audio etc etc)
• I implemented the PR with a lazy import - just checking that's in line with the repo's design/ vibe?