Hashmaps are used in Ndb for mapping a table rows hash value to a partition.

For fully replicated tables the implementation had two flaws:
1) It chosed a hashmap based on the number of fragments not the number of
   partitions.  For other kinds of tables these number is the same.
2) It did not actually use the hashmap as a hashmap, but rather as a partition
   to partition map, using the table rows hash value module parition count as
   input.

This patch fixes the above.