xmlrpc: Don't load all Patches into memory

xmlrpc: Don't load all Patches into memory

The way that reverse indexing of patches was implemented is broken.
At present, it will retrieve all patches in memory and return the
length from that data, then the slicing operation will then happen
without querying the DB and slice the results cached from the len()
evaluation. This is memory intensive, particularly for larger
instances.

Take advantage of Django's lazy loading to avoid this.

Signed-off-by: Stephen Finucane <stephen.finucane@intel.com>
Suggested-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Andy Doan <andy.doan@linaro.org>