Improve frequency handling in frf.build_fringe_rate_profiles

[r-pascua]

There have been multiple cases of users having unexpected results crop up when using frf.build_fringe_rate_profiles which ultimately trace back to how the frequency axis is handled within the function. The way in which the frequency axis is incorrectly handled is complicated to explain, but it is prone to silent bugs that are difficult to debug without going over the function's code line-by-line. This function's code should be cleaned up to remedy this.

[jsdillon]

I'll just add this link to the relevant slack chat for posterity: https://eoranalysis.slack.com/archives/C03DK432S4S/p1699041990832489