Indexing in-memory documents

[fschlatt]

I wanted to index a collection from in-memory documents. The collection loader already supports passing in-memory documents, the only issue was that the config would store and save the entire collection as well. To prevent this, I added a flag to the index function which allows specifying a collection name. The collection name is then saved instead of the in-memory collection. To make searching work, it was now also necessary to remove loading the collection before searching (since the collection config option now isn't a path but the collection name). As far as I can tell, loading the entire collection before searching is completely unecessary anyway and searching works perfectly fine without loading the collection beforehand.