Historical backfill performance improvements

[morgsmccauley]

The historical backfill process will first gather all related block heights, and then push them to Redis. The former can take 10+ minutes, causing a large delay between publishing the Indexer, and actually seeing Historical block executions.

Redis messages should be published as soon as they are available so that historical execution can happen as soon as possible.

Additionally, we should investigate the use of near-lake-framework within the Historical thread. Currently, we fetch then process each block sequentially. Using Near Lake framework will allow us to pre-fetch X amount of blocks ahead, allowing us to speed up this portion.