Take component size into account for Q13(a) and Q14(a) variants

[szarnyasg]

To make the queries really challenging, in the paramgen we could select components that are roughly similar in size -- and reasonably large, if possible.

[szarnyasg]

Getting two large components may not be possible... E.g. in the SF10,000 data set, most people are in a single large CC while the rest are all isolated nodes.

D create or replace table cc as select * from read_parquet('factors-sf10000/parquet/raw/composite-merged-fk/personKnowsPersonConnected/*.parquet');
D select distinct component, count from cc order by count desc limit 5;
┌────────────────┬──────────┐
│   Component    │  count   │
├────────────────┼──────────┤
│ 14             │ 26519261 │
│ 10995121104613 │ 1        │
│ 10995121125221 │ 1        │
│ 10995121148539 │ 1        │
│ 10995121161903 │ 1        │
└────────────────┴──────────┘

[szarnyasg]

See the related issue for the BI workload.