WL#1121: Simplifying test data generation

Currently to create large enough set of data we have to write the scripts
etc which makes generation of test cases/bugs hard.

As soon as we get INFORMATION_SCHEMA or similar virtual system tables we shall
also add virtual table which would allow us to generate dump data (sort of
/dev/zero) 

So we could for example use the following query to generate table with 
1.000.000 random values:

INSERT INTO random  select rand() from zero_source limit 1000000