feat: refactor loading and attaching data sources, and upgrade to llama-index==0.10.6

[sarahwooders]

Please describe the purpose of this pull request.
Currently, you can only load data sources from the CLI, and only attached data sources from the CLI and via /attach. This PR makes it so that passages can be programmatically inserted into a data source, and also loaded into agents.

This PR makes changes to make it easier to build the REST servers on top of the attach/loading code:

• Define an extensible DataConnector object for connecting to external data sources
• Upgrade to llama_index==0.10.6 and resolve breaking changes (note: migration will no longer work, since the previously saved vector indices are no longer compatible with the most recent version of llama index)

TODOS

• Create common base function that are called from both CLI and server:
  • Creating a data source
  • Parsing files into passages (other data sources types are a TODO)
  • Inserting passages into the data source
  • Attaching data to an agent

TODOs for next PR:

• Support data loading/attaching from REST API
• Implement File API for OpenAI-compatible assistants API

Related issues or PRs
#892

[cpacker]

rest ez