Estimate tip frequencies up to the month of latest available strains

[huddlej]

We currently estimate tip frequencies up to the current date of a given build, but the latest available sequences are often collected one or two months prior to the current date. As a result, the frequencies misrepresent what we know about the present. For H3N2, this lack of data also makes the forecasts less reliable because they end up being 14-15 month forecasts instead of 12-month forecasts.

Instead of using the current date as the max date for tip frequencies, we should dynamically determine the max date from the available sequences in the analysis, rounding up to the first day of the next month. For example, if the latest available sequence is from April 15, we would set the max date to May 1.