Template m1 m2 ordering fix

We noticed that some GraceDB events had m_2 > m_1, and our code was allowing that through, including when computing q = m_2 / m_1 for the prior bounds calculation, potentially causing some issues. This MR swaps the two parameters in that scenario, emitting a warning.

A test run shows the recovered parameter closer to the peak in the marginal posterior plots, as expected, though no major changes to the posterior itself.