Add API endpoint to export accuracy metrics from user feedback + created_at timestamp

[tholor]

User feedback collected at runtime can be utilized for two use case:
a) fine-tuning models on this custom training data (see https://haystack.deepset.ai/docs/latest/domain_adaptationmd)
b) evaluating models in production (i.e. creating accuracy metrics from the user feedback)

This PR implements basic functionality for b) by adding a new API endpoint /eval-doc-qa-feedback. When calling this endpoint we calculate simple accuracy metrics for answer and document results based on the given user feedback. Optionally, you can also supply filters to narrow down the labels that shall be used:

Example Request:

curl --location --request POST 'http://127.0.0.1:8000/eval-doc-qa-feedback' \
--header 'Content-Type: application/json' \
--data-raw '{
    "filters": {"document_id": ["a52c4f5b-2221-4653-8dc8-31eee2cf0ac4"]}
}'

Example Response:

{
    "answer_accuracy": 0.0,
    "document_accuracy": 0.0,
    "n_feedback": 1
}

Breaking change:

The labels have now a field called created_at that includes an automatic timestamp.

Before:

• In SQLDocumentStore and FAISSDocumentStore all objects had a field / column created
• In ElasticsearchDocumentStore labels did not have a field created

After:

• Labels in SQLDocumentStore, FAISSDocumentStore and ElasticsearchDocumentStore all have a field created_at