allow user to supply vector of tau, one value per colony

[Aariq]

I need to look more into the current behavior and think about desired behavior, but here's the general idea...

You may want to use a different set of covariates when estimating the switchpoint, tau, than when you estimate colony growth, given a switchpoint. The only way to really do that currently is to run bumbl() once, then extract the values of tau, then sort of manually fit the switchpoint model by calculating the .post covariate. If you could override the grid-search portion of bumbl by providing known taus, you could solve this issue by running bumbl() once to estimate tau for each colony, then run it again with bumbl(...tau = tau) and a different set of co-variates.

I think this would do what Elizabeth and Sylvie were hoping it would do.

[Aariq]

I no longer think this behavior is desirable.

[Aariq]

Actually, this seems like it could be a useful option still.