Save intermediate hazard curves to DB in UHS calculations

[Lars Butler]

[et=8h]
[at=]

Requested by Laurentiu Danciu:

In the Disagg and UHS calculators, we're generating hazard curves as intermediate artifacts. It would be useful for the users to save those intermediate results and output them from the calculation.

----------

To implement this, the following needs to be done.

Java side:
- Refactor the way the UHS Calculator (on the java side) constructs and returns results (see: https://github.com/gem/openquake/blob/ce4dbc1f83bca744df6bd66a2ecd946f807414ba/java/org/gem/calc/UHSCalculator.java#L121, https://github.com/gem/openquake/blob/ce4dbc1f83bca744df6bd66a2ecd946f807414ba/java/org/gem/calc/UHSResult.java). The result object needs to store the uhs (array) and poe pairs, as well as all of the hazard curves for the site.

Python side:
- Grab the hazard curves from the java result object and save them to the db. This new functionality should probably added to this method: https://github.com/gem/openquake/blob/ce4dbc1f83bca744df6bd66a2ecd946f807414ba/openquake/calculators/hazard/uhs/core.py#L163
- Add QA test coverage

DB:
- Include an optional sa_period attribute for each hazard curve node.
- Add `imt` and `imls` fields to the hzrdr.hazard_curve table. This information is available in the uiapi.oq_job_profile, but we have to do several joins to get this information. It would be more efficient (especially for export hazard curves from the DB) if we denormalized a bit and just stored a copy of this data in the hzrdr.hazard_curve table.

NRML:
- Include an optional sa_period attribute for each hazard curve node.

[Damiano Monelli (monelli)]

This is also useful because with a single UHS calculation, we can compute hazard curves at different periods with a single OpenQuake run (and so deriving hazard maps for different periods too), while with the classical PSHA calculator (as it stands now) we can compute hazard curves for only one period at the time.

[Lars Butler (lars-butler)]

@Damiano:

For each site of interest, we compute 1 hazard curve per SA period. When we store the hazard curves in the database/NRML, do we need the associated period as well? I would think that the answer is 'yes'; otherwise, there really isn't enough information to interpret the results.

[Damiano Monelli (monelli)]

@lars-butler:
Yes I think it's a good idea. For each curve we should also give the corresponding period.