sql/parser: add support for aggregate functions

[petermattis]

Aggregate functions perform some operation on the result set for a query. The basics are avg, count, max, min and sum. A typical query might look like SELECT count(*) FROM foo. Note that the aggregate function might affect index selection. For example, SELECT MAX(timestamp) FROM foo might be satisfied via a single index read if there is an index on timestamp.

Conveniently, aggregate functions cannot be mixed with non-aggregate values. This is not valid: SELECT AVG(k), k FROM foo

[petermattis]

Aggregate functions to implement (initially):

• AVG(expression)
• COUNT(*)
• COUNT(expression)
• MAX(expression)
• MIN(expression)
• SUM(expression)

The initial implementation can be fairly dumb: if an aggregate function is detected add a new aggregateNode which evaluates the expression. Optimizing MIN and MAX should be done where an appropriate index is available. Trying to distribute the aggregate calculation will be left for later.