Update code for how individuals enroll in treatment sites

This follows a conditional logit specification (i.e. single choice equal to maximal latent utility, with coefficients constant across choices) that has already been specified.

Predictors include distance and some predictive factors (like observed student variables) with coefficients in common across options, there should be treatment specific fixed effects (centerd around zero) in drawing attendance.

No files or tests are named; start by locating the treatment-site enrollment code and the existing conditional logit specification. Done should match the stated model: common coefficients for distance and observed student predictors, treatment-specific fixed effects centered around zero, and attendance drawn from the maximal latent utility choice.

Written by the indexing model from the issue text.