Add corpus label for identifying groups of documents

[nenb]

This PR addresses some aspects of #256. There are 3 changes:

1. A corpus_id kwarg has been added to core components like the source_storages and the Chat class
2. source_storage tables are now identified by the corpus_id label rather than a chat_id label
3. All use of the prepared flag has been removed.

More details
A corpus_id label allows easy reference to embeddings that have already been computed. Removing the convention of using chat_id to name the tables means that embeddings are no longer coupled to chats and can be re-used. Removal of the prepared flag is necessary to prevent inconsistencies when re-using embeddings across chats (eg it would be inconsistent for the same corpus to have prepared values of True and False for different chats, which may occur for a variety of reasons)

Not in scope for this PR
The UI will also require some UX work before it can make proper use of this PR. For now, I have simply hardcoded the corpus_id label to be equal to the chat_id value in the UI section of the code. This should give functionally equivalent behaviour as to what currently exists.

[nenb]

Closing as it turns out that it is possible to implement similar functionality by designing a custom Document extension and a custom source_storage extension.