Add array:unique langlib function

[jclark]

This is defined for anydata[], works like the Unix uniq command i.e. it creates a new array that omits any member that is == to an earlier member.

The input array would not be required to be sorted: the implementation would be expected to use a hash table to remove duplicates (or could create a temporary sorted array and use that).

Query will eventually be able to do this using grouping and aggregation (#441), but this is a lightweight, simple solution.