Update on the Analysis Code

I've made some decent progress on implementing the analysis framework. You
can find the latest code on this branch:
https://github.com/googlefonts/PFE-analysis/commits/extended_analysis_output

I've got a bit more to implement and then I'll merge everything into the
master w3c repo.

Currently I have the following working:
- Defined a format for the input data
<https://github.com/googlefonts/PFE-analysis/blob/extended_analysis_output/analysis/page_view_sequence.proto>
.
- Defined a format for the analysis results
<https://github.com/googlefonts/PFE-analysis/blob/extended_analysis_output/analysis/result.proto>
.
- Defined an interface for a PFE method
<https://github.com/googlefonts/PFE-analysis/blob/extended_analysis_output/analysis/fake_pfe_method.py>.
To add simulation of a PFE method simply implement the interface. See
the "whole
font" pfe method
<https://github.com/googlefonts/PFE-analysis/blob/extended_analysis_output/analysis/pfe_methods/whole_font_pfe_method.py>
for an example of what is needed.
- Implemented the patch subset PFE method (a python wrapper
<https://github.com/googlefonts/PFE-analysis/blob/extended_analysis_output/patch_subset/py/patch_subset_method.py>
around the C++ code).
- Implemented the basic analysis, currently it can take a set of network
models, pfe methods, and an input data set and will output the total cost
for each (method, network model). Note: the cost function is not yet
implemented.
- Set up continuous integration tests running against linux and mac OSX.

Myles, since you'll be working on implementing a PFE method for the byte
range approach I recommend reviewing the expected interface for a PFE
method and the input data format. Those can be found here:
https://github.com/googlefonts/PFE-analysis/blob/extended_analysis_output/analysis/fake_pfe_method.py
 and
https://github.com/googlefonts/PFE-analysis/blob/extended_analysis_output/analysis/page_view_sequence.proto
.