TODO

ToDo items:
 - Add an option to create per-PID files to record data.  This may help if running
   multiple concurrent threads to avoid one thread overwriting data for another.  
   This looks like:
    * <filename>_header - header data
    * <filename>-<pid#1> - performance for process with PID....
   *** Note that the data then needs to be coalesced, probably with something like: 
       cat <filename>_header <filename>-* (note the underscore and dash distinction).
       
 - Move the file open to the location where the write occurs.  No need to hold the 
   file open the entire time the process runs.  This probably comes by default with 
   the previous option due to filename differences.
 
  - Create a simple tool to do basic analysis.  E.g.:
     * Graph usages over time using process start and end times.
     * Identify likely CPU bound, memory bound, or IO bound processes out of a list.