UniversalDopplerMetric: follow-up of unused `_j` variables in XLALComputeAtomsForFmetric()

@karl-wette In !1631 (merged) Duncan removed some lines computing the a_a_j, a_b_j, b_b_j variables in XLALComputeAtomsForFmetric() to fix unused-but-set compile errors. Looking at the code in detail (well, enough detail to see what's going on from a flow perspective, but not enough detail to understand what is actually supposed to be going on),

• the whole j block starting from https://git.ligo.org/lscsoft/lalsuite/-/blob/20fb8fe4ccfd80e3d422d5194d976e88c9233aa6/lalpulsar/lib/UniversalDopplerMetric.c#L1838 seems useless, as it just assigns temporary variables and does error checks on them, but doesn't use them for anything (this seems to be unchanged by the deletions from !1631 (merged), the remaining temporary variables just have reused names and hence don't trigger the compiler complaints)
• however, there are various gsl_matrix_set calls below it, and it may be that the _j terms were actually forgotten there, which would be bad. Can you reconstruct if they were ignored on purpose before, or just forgotten, and whether this would have any impact?

If you're not familiar enough with the details of these codes either, we can also check with Reinhard.