tidyclust loopy integration

[EmilHvitfeldt]

This PR is the first push to have tidyclust leverage the new tuning infrastructure.

This PR is very much just a DRAFT. I know that there are things that needs to be sorted out, and cleaned up. But it acts a starting point, and as an initial gauge of the amount of changes would be required

Note on Parameter Registries

• parsnip models: Use parsnip::get_from_env() which queries parsnip's internal environment
• tidyclust models: Use modelenv for registration, but parsnip doesn't see it
• Solution: get_param_info() handles NULL gracefully, defaulting has_submodel = FALSE

alternative solution is to import modelenv and have it swap between them depending on the model type. Also worth noting that tidyclust doesn't have any models that use the submodel trick.

This means tidyclust models won't benefit from submodel optimization (multipredict), but this is a minor performance consideration.