Fix time estimation

[leftmove]

Time estimation has been broken for a while.

It is either too slow or not functional whatsoever. Additonally, the "algorithmn" is a simple divide by two function that gives an incredibly rough "estimation." Since wallstretlocal has gone public, there are now enough statistics in the database to create a reasonable average time required using log counts.

• Make estimation faster
• Using estimated log count from counted stocks, create time estimation based on average taken from statistics collection

[chrisgarcia4158]

Potential average aggregation query.

db.statistics.aggregate(
   [
     {
       $group:
         {
           _id: "$type",
           avg: { $avg: { $divide: [ { $size: "$log.logs" }, "$log.time.elapsed" ] } }
         }
     }
   ]
)

[chrisgarcia4158]

Results for query.

{ "_id" : "create-historical", "avg" : 14.349005705254921 }
{ "_id" : "create-latest", "avg" : 17.308616566972628 }

Historical queries do about 14 per second and latest queries do about 17 per second. In order to create an accurate estimation, take stock count, multiply it by two, and add four. I've added some more accurate code to the add_query_log function, but until that collects enough statistics this will have to do.