Re-design primitives

[tholor]

Let's re-think what core data types / primitives we need in Haystack and how we want to structure them (attributes, methods, types ...).

As of now, we have in schema.py:

• Document
• Label
• MultiLabel

In addition, we use in the REST API:

• Answer
• Request
• Response

What we should do:

• Let's discuss if we need more primitives (in the past we thought about splitting Label into multiple classes for QA / Retrieval; we also want to have "standardized" input / output objects for pipeline nodes)
• Let's examine if the current primitives need any refactoring (e.g. attribute names, redundancies ...)
• Let's document all primitives

Related to:

• Will help to standardize input/outputs of pipelines and therefore enable validation of pipelines (Validation of search Pipeline graph #600)
• Will help to standardize request and response types in REST API for different pipeline types
• Might simplify evaluation of different pipelines types
• Might be needed to store labels for other pipeline types

[brandenchan]

Am very much in favor of this! Please keep me in the loop with progress so we can make sure we have this well documented

[tholor]

Implemented in #1398