refactor prep_data functions (& formulas.py)

[jburos]

In part inspired by feedback from @julia326 & @armish, making a list of unresolved code questions.

Within [survivalstan/formulas.py]:

• re: timepoint_id = 'id'

  are these safe defaults to have, and would they ever be different values? in general, there are a lot of string constant names throughout the code - is there an initial validation you can run to make sure the columns exist by the names you expect to see?

• re: self._type = "wide"

  I see you doing these wide/long checks throughout the code and it might make it harder to maintain/improve this code, if, for example, you have to add support for a new type for some peculiar reason (if that is unlikely, then just ignore this comment). Maybe these checks can benefit from the recent enum support in Python?

• re: Class Survdata(pd.DataFrame)

  Do you think would it make the code easier to read if you have SurvData as a base/utility class and WideSurvData and LongSurvData extending it to populate survival_type?

• re: def len(self):

  can this get called before memorize_finish()?