a few more ilabel problems

[kejohnso]

Testing ilabel more extensively, I found some more places where the functionality is working wrong. I created several tests that still include only one site, but extend to use more than one GMM per LT. The jobs use the subduction interface floating ruptures source for a site in the Seattle Basin. To summarise the issues:

1. The weights from the Default LT are always used, even if the GMMs used are correctly selected. If weights are the same and in the same order for two LTs, the results can be correct, but this is the only case I was able to find where results are correct.
2. When a DummyGMPE is used in the Default LT, rlz000 is all zeros, unless there is no ordinal defined in the LT branch in which case the zeros can correspond to a different rlz (note: I discovered this part later and don't yet understand it well).
3. If the DummyGMPE is in a non-default LT, the ordinal cannot be defined or the job will fail (but anyway the results are wrong).
4. I don't yet understand why case_06 in the qa_tests is OK.

I will share separately the test jobs and more thorough accounts of the problems.