851: Refactor data loaders and disable caching

[michael-markl]

Closes #851. Refactoring of the data loaders:

• Disable caching on all data loaders
• Make registering and naming of data loaders less boilerplaty and less error-prone
• Avoid name conflicts of data loaders
• Make them more type safe

Also, I rewrote the data loader for physicalStoreByStoreIdLoader which makes the searchAcceptingStores GraphQL query a lot faster: Instead of ~20 single SQL queries (one for each store) we now make a single SQL query.

[sarahsporck]

It seems you forgot sth. here.

Suggested change

	physicalStoreByStoreIdLoader.fromEnvironment(environment).load(id)
	physicalStoreByStoreIdLoader.fromEnvironment(environment).load(id).thenApply { it!! }

[michael-markl]

Actually not: We designed our API to support AcceptingStores that do not have a PhyiscalStore, (so this is a 0..1 connection). Hence, the physicalStore function here should return a nullable value.

[michael-markl]

In contrast to that, the above occurences of the data loader should always return a non-nullable field, as we require a category and a contact object on each accepting store.

(So previously the typing of the data loader was false in the above functions contact and category. That was also why I refactored it so that we don't have to specify the generics of the data loader here, but instead infer it directly from the dataloader we use)