Add capture of query load metric

[huitema]

We want to experiment with measurement of query load.

1. Query rate: average number of queries per time slice of 10 seconds
2. Standard deviation: std of the number of queries per 10 seconds, across the whole set
3. Max: max number of queries per 10 seconds, across the whole set.

The first step is to upgrade the capture program with a few new data points:

1. Number of 10 second slices in segment
2. Sum of queries received over all slices
3. Sum of squares of number of queries received over all slices
4. maximum number of queries per slice

We will also update the summary program, so the max is computed across all maxes, instead of just doing a summation.

Then we will update M3 to display the average, std and max -- i.e., the most simple part. We can then play with the data set and try find variations per locale, day, etc.

[huitema]

After discussion, we just did something much simpler. Count the total number of requests (and bytes) in a capture session, and the duration of the capture. This gives us unquestioned raw data. Then, we will try build a model.