Don't index blank timeseries

[corviday]

One of the test files we've been using has several completely masked timeseries. These aren't useful for anything, and shouldn't be indexed. The indexer should log a warning about them, and not put them in the database.

[corviday]

Apparently blank timeseries are an expected feature of datasets in the future, as there are some variable calculated only for lakes and some only for rivers. So if we have rivers and lakes in the same datafile, some of them will be blank.