As more endpoints started using the API response paginator, new patterns
started to emerge that required either more flexible configuration or a
lot of re-used code.

This commit splits the paginator into three logical components:

- The Paginator is reponsible for splitting the values of an iterator
  into chunks of the requested size.
- The Tokenizer returns a page token that is used to uniquely identify
  elements of the iterator.
- The Filter is used to decide if a given element should be added to the
  page or skipped altogether.

The Paginator implementation has returned, more or less, to its initial
clean design and implementation.

Additional logic and flow customization can be provided by implementing
the Tokenizer and Filter interfaces.

One Tokenizer implementation is provided (StructsTokenizer) which can be
configured to emit tokens using common fields found in structs, such as
ID, Namespace, and CreateIndex.

The go-bexpr Evaluator implements the Filter interface, so it can be
used directly. Another implementation provided is the NamespaceFilter,
which can be used to apply filtering logic based on the namespaces the
ACL token used for the request is allowed to access.

It's currently not possible to provide consistent sorting by CreateIndex
to this endpoint because it requires a Namespace and ID to be able to
uniquely identify a volume.

This reduces the type of go-memdb indexes that can be created where the
raft CreateIndex would be in a position that would allow sorting with
prefix match.

It's currently not possible to provide consistent sorting by CreateIndex
to this endpoint because it requires a Namespace and ID to be able to
uniquely identify a job.

This reduces the type of go-memdb indexes that can be created where the
raft CreateIndex would be in a position that would allow sorting with
prefix match.

Co-authored-by: Seth Hoenig <seth.a.hoenig@gmail.com>

Refactor the allocations list endpoint to merge the listAllNamespaces
special case into the normal List function.

Co-authored-by: Seth Hoenig <seth.a.hoenig@gmail.com>

Refactor the jobs list endpoint to merge the listAllNamespaces special
case into the normal List function.

Using reverse provides a more generic intention that the results should
be returned in the reverse of the default order instead of assuming any
particular order.