Generating histograms of large data

Continuing from where I left off at yesterday's histogram generation.  It's now running on a large data set. Even though it's a lot quicker to run the application now (15 minutes vs probably 15 days on the Opera loading google.com dataset), it's still a bit of a chore to regenerate each time.

Especially since the data is going to be used as input for my measurements of rmalloc, specifically when to insert lock/unlock operations.  Lifetime data is now cached as a json blob in a file to be used by other tools. The histogram tool checks the contents of that file, and if it can be loaded, it is used instead of the ops file.

Checking out after 3h5m,
Mikael.