Two bugs in dealing with datasets that cancel each other

[olaurino]

According to the Sherpa-Specview interface design, a client should be able to send a message (e.g. calc_model_values) with several datasets that can be simultaneously fitted or evaluated.

However, this does not work as expected, although the current Iris implementation never relies on this feature, so it does not impact this functionality.

In sherpa_samp/mtypes.py, line 890, it is assumed that there is only one dataset in the session, with id 0:
ui.session.get_data(0).x

Also, in sherpa_samp/session.py, method set_model (line 119), it looks like the models are applied to the wrong datasets, looping over the models instead of over the datasets. Again, as multiple datasets are never exercised in Iris, this bug doesn't have an impact, although it impacted the implementation of new features and required working around it.