Use cursors instead of loading entire result set into memory

[onbjerg]

Right now we do a single query to get every transaction and log in a block and load this directly into memory. However, some blocks contain very large transactions with a large number of logs, which might cause the worker to run out of memory.

Instead, we should use a cursor and iterate over rows 1-by-1.

Documentation

[onbjerg]

No longer relevant since we're moving to The Graph (#200).