add concept of execution time to diff

[sdmcclain]

This PR does two things:

1. Adds the concept of Execution time, by printing it to the top of the file.

In https://github.com/mcclaized/edav_final_project/blob/master/bin/run_pyflame, I run the code twice, first to grab the execution time of the program with no profilers adding overhead, the second using pyflame to generate the profile.

I am passing this forward, because it will be used in sdmcclain/node-stack-convert#1 to determine the number of seconds that each function call contributes to the overall runtime.

2. Adds a new mode (triggered with the -u option) that compares only the differences of two profiles. https://github.com/corpaul/flamegraphdiff deserves a major shout out for the inspiration of this work, but my implementation is somewhat different.

Without -u, each line represents a call stack, and is followed by 2 numbers, representing the number of times the two inputs files were sampled at this call stack:

difffolded.pl profile1.folded profile2.folded gives
my_file.py:38 4 2

where 4 is the number of times the program was sampled at line 38 of my_file.py in profile1 and 2 times in profile 2.

In the -u mode, the output is different. We first discard any case where the number of calls is the same, then we take the diff abs(profile2_calls - profile1_calls) and store it as the first number, with the sign as the second number. So, in the above example, the output would be:

my_file.py:38 2 -1

Because 2 is the difference in number of calls, and the -1 signifies that the second file has fewer calls than the first.

See sdmcclain/node-stack-convert#1 for details on how this will be used