Datasets: output corresponding type

**What's your use case?**

Datasets always outputs an instance of Table, even for data that would be more suited to Corpus.

**What's your proposed solution?**

Datasets could (should?) output Corpus when the data is more appropriate for text mining and Timeseries when appropriate for ts. A flag somewhere?

**Are there any alternative solutions?**
Currently, the user has to explicitly use Corpus and set text features manually.

The issue names Datasets, Table, Corpus, and Timeseries but no files or tests. Start by locating the Datasets implementation and the existing Corpus and Timeseries data types, then determine how their selection could be specified and tested. Done should be defined by documented behavior and tests showing appropriate output for tabular, text-mining, and time-series data.

Written by the indexing model from the issue text.