distributed Query Plan cache may be poisoned by experimental query plan features

[fotoetienne]

Describe the bug
If a router instance is sharing a distributed cache with other router instances, and one router instance used experimental query plan features, the resulting query plans will be cached with the same id as query plans that are not using the experimental features.

To Reproduce
Steps to reproduce the behavior:

1. Deploy a cluster of router instances sharing a distributed cache
2. Deploy a single instance of the router (sharing the same distributed cache) that is configured to use experimental query plan features such as generate_query_fragments
3. Issue a request against the new router instance and then the same query against the other instances
4. All instances will issue a query using the experimental query plan feature

Expected behavior
Query plans created with different federation versions and federation options will be cached with a unique key

[fotoetienne]

A proposed solution is to add QueryPlannerConfig to CachingQueryKey (PR). Thoughts?

[Geal]

yes, adding it to the cache key is the right approach here