test.BernoulliCountQuery : sampling rate in percentage. arg = 0.001 =>
sampling rate = 0.00001. n = 1B => sample size = 10000.


AB-tree's rejection rate hike example due to sampling from a small range
sitting between two adjacent sub-trees:

load the test table A, with test.InsertMany 100000000

bulk load an AB-tree on y

query: test=# select min(1/sample_prob()), max(1/sample_prob()),
cast(count(sample_prob()) as float8)/count(*) as acc_rate from A tablesample
swr (10000) where y between 110000000 and 110000300;
INFO:  N=10000
 min | max | acc_rate
 -----+-----+----------
  783 | 783 |    0.381

(1 row)