Add better null handling to checks about tattle HOT 2 CLOSED

In the branch I'm working on, I use Graphite's movingAverage function to smooth out the data. I use the Check's sample size to plug into the movingAverage's 'number of data points' argument and threw out Tattles calculation of the average. I instead use the last data point that is returned in the JSON object (which has the movingAverage built into the results) when I call getData and compare it against the threshold to determine the status of the Check. (The only downside to this is I can't use median as the baseline anymore)

In one of my cases, I have a chron job I track in Graphite that runs every 30 minutes, where each time the job should run against approximately 1/3 of my user base. (I therefore have approximately 29 out of 30 data points that are returned as 0) I use a sample size of 90 that smooths out my data pretty well. Even with the movingAverage smoothing out the data on a 90 minute basis, I get alerted when 'all the values are null' usually within the hour. Overall, I disagree that 'we should probably ignore nulls' just because the data isn't there. The data is still a representation of events that are (not) occurring, and should be accounted for as such.

from tattle.

@ssandler I don't think that Tattle should ignore nulls in every case, maybe it could be a checkbox or whatever option, if you wan't to do such a pull request, but anyway graphite's transformNull() function can do the job as well (see http://graphite.readthedocs.org/en/latest/functions.html#graphite.render.functions.transformNull)

from tattle.