feat(parachain/availability-distribution): implement management of chunk fetching tasks

[haikoschol]

Issue summary

This functionality is part of the message handling logic for ActiveLeavesUpdateSignal.

Implementation details

For each newly activated leaf plus the ancestors retrieved in the previous step, instantiate a runtime instance and call ParachainHostAvailabilityCores(). Filter out all cores that are unoccupied or not associated with the relay parent. For each core, check whether there is already a fetch task running for the candidate hash associated with the core. If yes, add the newly activated leaf to the state of the fetch task.

If not, determine the chunk index for the core and start a new fetch task for this chunk.

For each deactivated leaf, remove it from the state of all running fetch tasks. Cancel all fetch tasks that have an empty list of leaves.

package availabilitydistribution

type BlockState interface {
    GetHeader(hash common.Hash) (*types.Header, error)
    GetRuntime(hash common.Hash) (instance runtime.Instance, err error)
}

type fetchChunkTask struct {
    leaves []common.Hash
    // ...
}

type AvailabilityDistribution struct {
    blockState BlockState
    fetches map[parachaintypes.CandidateHash]*fetchChunkTask
    // ...
}

Other information and links

• get_occupied_cores()
• add_cores()
• stop_requesting_chunks()

Acceptance criteria

• New code is 60% covered with unit tests