Avoid querying the local validator in availability recovery (#2792)

* guide: don't request availability data from ourselves

* add QueryAllChunks message

* implement QueryAllChunks

* remove unused relay_parent from StoreChunk

* test QueryAllChunks

* fast paths make short roads

* test early exit behavior

polkadot/node/network/availability-distribution/src/requester/fetch_task/mod.rs

@@ -27,7 +27,7 @@ use polkadot_node_network_protocol::request_response::{
 	v1::{ChunkFetchingRequest, ChunkFetchingResponse},
 };
 use polkadot_primitives::v1::{AuthorityDiscoveryId, BlakeTwo256, CandidateHash, GroupIndex, Hash, HashT, OccupiedCore, SessionIndex};
-use polkadot_node_primitives::ErasureChunk; 
+use polkadot_node_primitives::ErasureChunk;
 use polkadot_subsystem::messages::{
 	AllMessages, AvailabilityStoreMessage, NetworkBridgeMessage, IfDisconnected,
 };
@@ -405,7 +405,6 @@ impl RunningTask {
 			.send(FromFetchTask::Message(AllMessages::AvailabilityStore(
 				AvailabilityStoreMessage::StoreChunk {
 					candidate_hash: self.request.candidate_hash,
-					relay_parent: self.relay_parent,
 					chunk,
 					tx,
 				},

polkadot/node/network/availability-distribution/src/tests/state.rs

@@ -67,7 +67,7 @@ pub struct TestState {
 	pub relay_chain: Vec<Hash>,
 	/// Whenever the subsystem tries to fetch an erasure chunk one item of the given vec will be
 	/// popped. So you can experiment with serving invalid chunks or no chunks on request and see
-	/// whether the subystem still succeds with its goal. 
+	/// whether the subystem still succeds with its goal.
 	pub chunks: HashMap<(CandidateHash, ValidatorIndex), Vec<Option<ErasureChunk>>>,
 	/// All chunks that are valid and should be accepted.
 	pub valid_chunks: HashSet<(CandidateHash, ValidatorIndex)>,