create_id_breakdown is a bottleneck and needs rewriting

[ross-spencer]

Given an 8 million line SF YAML, (631,286 row database), create_id_breakdown is taking too long. It is largely unoptimized and not brilliantly written. Any rewrite I believe should bring pretty decent efficiency gains. Lets have a look at what we can do.

Edit: For reference, without this function alone, the script is quicker by over an hour, and completes in 77 seconds. There may be other bottlenecks along the way as much relies on the output here, but one step at a time.

NB. Rewrite could be focused on better sqlite queries which do not seem to be a bottleneck at all. Or it could be focused on improving the data structures we're using.

[ross-spencer]

There's a spare index described here: exponential-decay/sqlitefid#9 that might be worth looking into for performance.