Refactor aggregate context to be carried with the scope.

[scudette]

1. Each aggregate function instance in the AST carries a unique
   ID. This allows the function to store its own state in the
   aggregate context without interference from other instances of
   the same function (e.g. having two count() instaces is OK)

2. The scope may contain a reference to an AggregatorCtx
   object. This object manages access to the aggregate
   context. The main method that should be used is Modify() which
   mofidies the context under lock.

3. When the scope spawns a child scope, the child scope does not
   have its own AggregatorCtx, instead chasing its parent to find
   one. This allows aggregate functions within the scope to see
   the wider AggregatorCtx which controls the entire query clause.

4. When the query runs in an isolated context, the AggregatorCtx
   is recreated at the calling scope. This allows isolated scopes
   to reset the AggregatorCtx. For example, when calling a LET
   defined function, a new context is created.

5. If a GROUP BY query, the Grouper will create a new
   AggregatorCtx for each bin. This allows aggregate functions to
   apply on each group separately.