Add compare_target_qa script

[sbailey]

For consideration, this PR adds a compare_target_qa script for facilitating side-by-side comparisons of run_target_qa outputs, e.g. for comparing the same code on different datasets or different target selections on the same dataset. It has some hardcoded assumptions about some of the files to expect and a naming scheme of {directory}/{prefix}-{targettype}.png, but it does have some mild robustness to additional files being added or missing files from one run to another.

Examples outputs are at:

• Comparing DR4 to DR6 with desitarget/0.19.0: http://portal.nersc.gov/project/desi/users/sjbailey/targetQA/compare/dr4_0.19.0-dr6_0.19.0.html
• Comparing DR4 to the simulations in the 18.2a reference run:
  http://portal.nersc.gov/project/desi/users/sjbailey/targetQA/compare/dr4_0.19.0-18.2a_0.19.1.html

Note: these comparisons aren't taken at random; they have some distinct and important differences.

For running at NERSC, it requires the input QA directories to already be web visible because it provides relative paths to those files; it doesn't copy the images into a new location. If the original QA directories are deleted, it will break the outputs of the comparison page too.

Contributed improvements are welcome, but I'm not likely to spend much time personally on polishing this. Take it and run with it if you like, or if @geordie666 / @moustakas find this approach fundamentally flawed or unhelpful, we can also toss this PR and I can keep the code for my own private use. But it seems we want something like this to facilitate QA comparisons without toggling between a bunch of tabs and links.

[geordie666]

This seems to be a convenient standalone script. I see no particular reason not to include it, whether or not we build on it or maintain it, so I'm merging.