Context:

As discussed in this PR comment: https://github.com/camunda/camunda-optimize/pull/3245#discussion_r592146241 

Currently, when importing data for indices which might not exist (eg: decision, process or event process instance data), the writers call a method which checks whether the required index exists and, if it doesn't, creates a new index. As mentioned on the PR, the fact that these writer methods not only write new data but also potentially create new indices is not obvious, it would be less misleading to have the respective import jobs separately call the index check/creation method and then the data writing method.
Let's adjust this pattern in our code base where it appears.

AT:

• generateImports methods no longer create indices
• indices are created separately before the respective generateImports methods are called