Refactor execution state reads

[damip]

Closes #4075

• removes a few deadlock risks
• adds efficient prefix search inside datastores
• adds the query_state atomic state query system to Execution
• improves db queries

[aoudiamoncef]

@damip I have a question about the bindings of the responses to the initial request filters:

Currently, we are just using the order, with this way, we make things simple, but we close the door to parallelisation ?

[damip]

@damip I have a question about the bindings of the responses to the initial request filters:

Currently, we are just using the order, with this way, we make things simple, but we close the door to parallelisation ?

Actually we can already parallelize it with the order approach: just run all the requests (they are all read-locking) in parallel, see rayon, then gather all the results and return them in the right order

[damip]

@aoudiamoncef @Leo-Besancon ready for review + merge

[Leo-Besancon]

Generally a good refactor! I didn't catch anything critical, but I think during our labnet tests we should check that the pos_state is consistant.

[aoudiamoncef]

in API side, LGTM

[damip]

MACOS fails unrelated bootstrap bandwidth tests. I think we can merge