perf

Test the perf of some Markdown implementations.

Generate some test cases:
```
 ./gen_perf_cases.py [limit]
```
for example:
```
 ./gen_perf_cases.py 1000
```
This created a bunch of (small) test .txt files in "cases". These are derived from a bunch of Python Cookbook data. "limit" is a max number of "recipes" in the data set for which to generate cases.

The test files are small and don't necessarily a lot of markup, so this may not really be a good breadth perf suite -- it is real data though.
Process the Markdown for each "cases/*.txt" with Markdown.pl, markdown.py and markdown2.py:
```
 ./perf.py
```

strip out the .text cases that markdown.py blows up on? (encoding problems)
add some larger perf suites (perhaps those test case files that all Markdown implementations pass)
add markdown.php timing