* A single ChunksCoordinator is now in charge of following a shard and
keeps on coordinating until the persistent task has been stopped. Whereas
before a ChunksCoordinator's job was to process a finite amount of chunks
and then a new ChunksCoordinator instance would process the next chunks.
* Instead of consuming the chunks queue and waiting for all workers to
complete, another background thread will continuously and chunks to the
queue, so that the workers never run out of chunks to process if the
leader shard has unprocessed write operations.

Relates to elastic#30086