Lazy retrieval of Schemas

This is affecting the dev-v3 branch only.

We currently get a set of schemas, and one has to work to make the validation succeed. This means, in the worst case, we have one HTTP request too much. This is currently not an issue. But if retrieving/creating a schema would get more expensive, or we support more types of schemas, this overhead will grow.

A solution would be some kind of lazy retrieval of schemas. Only get the second, if the first did not succeed with the given value.

The issue names only the dev-v3 branch, so start by tracing where schemas are retrieved and validation is attempted, then inspect the current sequence of HTTP requests. Done means the second schema is retrieved only when validation with the first schema fails, without changing successful validation behavior.

Written by the indexing model from the issue text.