Don't use the lifelines implementation, just implement it ourselves.
Turns out the output of the nonparametric model and Kaplan-Meier is basically identical. I'm going to do some more testing but I'm tempted to delete the former given that it's a lot more complex/slow
Coverage increased (+5.04%) to 91.688% when pulling cbdc979fd7474ab3c8c53f9865bde7d55291e843 on rewrite-kaplan-meier into 25ee561591badde956749864ae12ba11d9d7ff2e on master.
Don't use the lifelines implementation, just implement it ourselves.
Turns out the output of the nonparametric model and Kaplan-Meier is basically identical. I'm going to do some more testing but I'm tempted to delete the former given that it's a lot more complex/slow