Better bundling algorithm; change file order if possible.

Currently playing around with this query

SELECT 
  count(advagg_aggregates.filename_hash) AS counter,
  advagg_aggregates.porder - a.porder AS location,
  advagg_aggregates.porder,
  advagg_files.filename,
  advagg_files.changes
FROM `advagg_aggregates` 
INNER JOIN `advagg_files`
  ON advagg_files.filename_hash = advagg_aggregates.filename_hash
INNER JOIN (
  SELECT 
    advagg_aggregates.porder,
    advagg_aggregates.aggregate_filenames_hash
  FROM `advagg_aggregates` 
  INNER JOIN `advagg_files`
    ON advagg_files.filename_hash = advagg_aggregates.filename_hash
  INNER JOIN `advagg_aggregates_versions`
    ON advagg_aggregates_versions.aggregate_filenames_hash = advagg_aggregates.aggregate_filenames_hash
    AND root = 1
  -- WHERE advagg_aggregates.filename_hash = 'W542ihfXAG6QdAnSs1yG4Whdd7pYpjl0EW5daJ3EQBw'
  -- WHERE advagg_aggregates.filename_hash = 'vNguyKT44qzCV0N-b6VsRqkKDuhES_FNB5ugGBriTwg'
  WHERE advagg_aggregates.filename_hash = 'iUVyBLVVmAsFWnYi5FAG4DrjaZoffblUg4Yqt4cBKRI'

) a 
  ON a.aggregate_filenames_hash = advagg_aggregates.aggregate_filenames_hash
 -- AND advagg_aggregates.porder BETWEEN a.porder - 3 AND a.porder + 10
GROUP BY advagg_aggregates.filename_hash
ORDER BY 
  counter DESC, 
  location ASC

Using this it will identify what files have a high location but a low counter. These are candidates for moving to a later location; thus giving better bundles.