Decide if the leap second wrapper is the long-term API

[eteq]

As discussed in #49 (comment) and implemented in #58, there is now an API for updating the leap seconds. However, some compromises were made in this API, as it was basically done using Astropy as a "guinea pig". So this issue is a space to discuss practical drawbacks of the current API and would might be done to either improve or replace it.

There's not a clear end-point here, but tentatively I'd say if in 2 years time (i.e., the end of the LTS cycle of the Astropy version that this is used in), if we are happy with it we should declare the API "good". So by end of 2021 this can be closed if not before.