Make ds9 region file parser fast

The docs contain this note: "pyregion is rather slow, likely due to a inefficient parser."
( see http://pyregion.readthedocs.org/en/latest/index.html#documentation )

It's really slow ... parsing 1000 simple circle regions

``` python
with open('ds9.reg', 'w') as f:
for _ in range(1000):
f.write('circle(0,0,1)\n')
```

takes 7 seconds:

```
%time pyregion.open('ds9.reg')
CPU times: user 7.59 s, sys: 20 ms, total: 7.61 s
```

Astropy bundles [ply](http://www.dabeaz.com/ply/), so if someone takes this on it might be worth considering using `ply` (currently [pyparsing](http://pyregion.readthedocs.org/en/latest/users/installing.html#pyparsing) is used).

I'm using `pyregion` in a data analysis / image plotting pipeline and region file parsing is actually the bottleneck, but it's acceptable and I'm not very familiar with parsing and region files, so I don't plan to take this on ... just wanted to make a GH issue so this known issue is not forgotten.

Start at pyregion.open and the current pyparsing-based parser, then reproduce the 1,000-circle benchmark described in the issue. Done means materially faster parsing of that example while preserving existing behavior; no implementation path or test file is named, so parser design and regression coverage need to be established.

Written by the indexing model from the issue text.