Add BTree Indexing to File Backend

[adewes]

Currently the file-based backend uses a simple hash map to store indexes on disk. This has the drawback that loading and storing the index becomes very expensive when many documents are in a given collection.

Help us to improve the file backend by adding a BTree-based index:

• Create a subclass derived from TransactionalIndex
• Redefine the add_key and remove_key so that they use a BTree to instantly store keys in the index

This is a complex task, please feel free to ask questions and post comments and ideas in this thread!