fix handling of horizons for sarix and gbqr models

[elray1]

Fixing a behavior of the outputs:

• previously, horizons of the output were relative to the last observed data point, which was not in alignment with the flusight format
• now, horizons are relative to the reference date, as expected for flusight submissions

This required changes to the code, to fix the computations, as well as to the expected outputs for integration tests. Basically, I added 1 to all horizons in the expected output data files.

The integration tests pass for me locally:

=========================================================================== 3 passed, 21 warnings in 66.84s (0:01:06) ===========================================================================

[matthewcornell]

approved via live meeting