API: Calculate the mean Brier score of a forecast

Best would be averaged over time instead of just the mean of all estimates, otherwise making a lot of tiny updates just before resolution would heavily skew the mean.

Start by locating the API entry point that calculates Brier scores and trace how forecast estimates and resolution times are represented. Define how the score should be weighted across time, including updates made shortly before resolution, then add coverage for forecasts with uneven update timing. Done means the API reports the intended time-averaged Brier score.

Written by the indexing model from the issue text.