Fix the benchmarks

[shama]

The benchmarks have fallen out and should be updated and fixed to compare the impact of each commit.

[rgaskill]

The benchmark test is missing changes because it is updating the file within a second of the timestamp on the file being changed. The OS X filesystem timestamp only has a time resolution of one second. So, any updates that happen within a second of the timestamp on the file will be missed by fs.watchFile because it sees no difference in the file timestamp (mtime).

This means the minimum time between changes that can be monitored is one second, but any file with a timestamp older than 1sec the change can be detected as fast as the poll cycle is. Does that make sense? It is a subtle distinction.

In the context of this benchmark test, one needs to wait at least a second after the creation of the test file and at least a second between each modification of the file or the change won't be detected.