Description
-----------

Histogram support was provided by WL#8705 but this was without any support
from InnoDB. The sampling was done by reading the entire table regardless of
the sampling percentage which meant that for large tables the sampling was
costly even when only a small part of the table needed to be sampled.

This worklog provides the required support from InnoDB to read only the
required sampling percentage of data. For ex, for SYSTEM sampling (which is
the only supported type of sampling currently), if the required sampling
percentage is 10, then we'd read (all the records in the page) psuedo-randomly
only 10 leaf pages if there are about 100 leaf pages present in the table
ensuring in the process that the data is evenly distributed and the sampling
is deterministic.

The worklog also introduces two counters sampled_pages_read and
sampled_pages_skipped in the module "sampling" to track the number of pages
read and skipped during the sampling.

RB#: 22014
Reviewed-by: Annamalai Gursusami <annamalai.gurusami@oracle.com>
Reviewed-by: Mayank Prasad <mayank.prasad@oracle.com>