dives if there are many ranges

For queries of the type

SELECT * FROM t1 WHERE a IN(<list of values>)

the range optimizer will estimate the cost of reading rows 
that qualify. There are two possibilities: 
 1) If there is a unique index on 'a', the row estimate is 1
 2) Otherwise, two dives into the index, one at range start 
    and one at range end, are done and the number of rows in 
    the range is used as estimate.

However, when there are a great number of values in IN(), the
time used to estimate the number of rows may easily exceed the
time it takes to execute the query. This WL is for reducing 
query optimization time by replacing one index inspection per 
range with already calculated index statistics for equality 
ranges. 

Since the index statistics are faster in use but less accurate 
than inspecting the index, different users may have different 
criteria for when to use which. A SESSION variable 
(eq-range-index-dive-limit) is therefore added as well: If 
the number of equality ranges is equal to or higher than this
value, index statistics is used for the equality ranges. The 
default is that index statistics is used if there are 10 or 
more equality ranges.