Storage of GMF Scenario IMT information is inconsistent

[Lars Butler]

The storage of IMT information for `GmfScenario` results (https://github.com/gem/oq-engine/blob/9b17fe4f7da998716ae8d58eaa897b8d6c350afa/openquake/engine/db/models.py#L2172) is inconsistent with other artifacts, such as hazard curves (https://github.com/gem/oq-engine/blob/9b17fe4f7da998716ae8d58eaa897b8d6c350afa/openquake/engine/db/models.py#L1751).

This causes results exported in XML format to be slightly wrong.

We get this:
<nrml xmlns:gml="http://www.opengis.net/gml" xmlns="http://openquake.org/xmlns/nrml/0.4">
  <gmfSet>
    <gmf IMT="SA(0.1)">
      <node gmv="1.11323674151e-24" lon="0.0" lat="0.1"/>
      <node gmv="1.36015602637e-24" lon="0.0" lat="0.0"/>
      <node gmv="1.37733022435e-24" lon="0.0" lat="0.2"/>
    </gmf>
...

We expect this:
<nrml xmlns:gml="http://www.opengis.net/gml" xmlns="http://openquake.org/xmlns/nrml/0.4">
  <gmfSet>
    <gmf IMT="SA" saPeriod="0.1" saDamping="5.0">
      <node gmv="1.11323674151e-24" lon="0.0" lat="0.1"/>
      <node gmv="1.36015602637e-24" lon="0.0" lat="0.0"/>
      <node gmv="1.37733022435e-24" lon="0.0" lat="0.2"/>
    </gmf>
...

[Michele Simionato (michele-simionato)]

There is an easy fix in the function openquake.engine.db.models.get_gmfs_scenario. I have also added in the export tests a check to make sure that the generate XML is valid for the NRML schema. See https://github.com/gem/oq-engine/pull/1132