Region aware storage version lag

Status currently computes version lag by subtracting its version from the log's version that is sent to it when peeking mutations. Because one potential source of lag is that the storage server is failing to peek mutations, we also separately compute the lag by subtracting the storage version from the largest log version and take the larger of the two computed lags.

Because the second lag is computed without regard for where the logs are located in multi-region configurations, this can lead to overreporting of lag on the storage servers.

Start at the Status code that computes version lag while peeking mutations, then trace how the largest log version is selected in multi-region configurations. Done means storage-server lag is not overreported when logs are located in other regions, while still accounting for failures to peek mutations.

Written by the indexing model from the issue text.