fix: optimize memory for stacks tsv import into rocksdb

[rafaelcr]

This PR changes the way chainhook imports a Stacks node TSV into rocksdb.

Before, it loaded the entire canonical chinstate (including the full block JSON messages) onto a VecDeque in memory and then drained that data into rocksdb. This was a very memory intensive process which crashed our dev pods ever time it ran.

Now, the process was changed to a VecDeque that only keeps the line numbers of the TSV where the block data exists, so it can later read blocks from the file 1 by 1 and insert them into rocksdb.

[tippenein]

Does this have a noticeable difference in processing speed? Does that even matter?

Code looks good. Just curious if you've run this

[rafaelcr]

Thanks @tippenein . Yep, it does, it runs a bit faster I assume because it doesn't have to deal with the very large data structure that was in place before but perhaps that's only in the dev env I was using to test