Modify siphon example to illustrate indexing and looping over datasets

[deeplycloudy]

Fixes #117. Adds an example of making the wrong assumption about how cat.datasets works as part of teaching the API.

Also changes the description of how filtering by time works to indicate that the filename has to match the built-in pattern. I think that behavior is provided by the default regex, not THREDDS, but didn't verify.

[dopplershift]

Yeah, THREDDS isn't doing any of that filtering, that's all in the Python.