non-uniform lead/lag effect of holidays in Prophet

Hi,
Section 3 of the [paper](https://peerj.com/preprints/3190/) explained how the effect of holidays can be modeled. The process is very similar to tuning the parameters in a regression for nominal values (non-ordinal categorical values using hot-encoder).
Consider a holiday such as Thanksgiving and assume our goal is to predict flight ticket sales. Intuitively, people want to be with their family on Thanksgiving day. Therefore, we expect a gradual increase in sales up to the day before thanksgiving (a very exaggerated figure is attached.) How *Prophet* can take that into consideration? In other words, when the effect of a holiday is non-uniformly distributed before and after the actual date, but not on that date, how should we approach the problem using *Prophet*?
Thank you