Add station and elevation dependence for coherence time

[dpesce]

Currently the atmospheric coherence time is assumed to be constant across the entire array, and it also does not vary with source elevation. This issue stems at least in part from an ambiguity between "coherence time" and "SNR threshold integration time," which are essentially conflated right now in the code.

An elevation dependence along the lines of that used by the ARIS (Asaki et al. 2007) or MeqSilhouette (Natarajan et al. 2022) softwares would be straightforward to implement, but would involve re-interpreting the user-specified timescale as some sort of coherence time. Because that would break backwards compatibility, a full addressing of this issue will probably require incrementing the major version number of the codebase.