Ratekeeper failing to limit on low TLog space

A post [on the forums](https://forums.foundationdb.org/t/foundationdb-seems-keep-reinitialising-itself-after-client-hang/1437/) shows a database that was permanently down due to transaction logs getting an error when trying to extend the disk queue. This implies that something failed between Ratekeeper and TLogs, because FDB shouldn't have been allowed to get into a situation where transactions are being committed that we don't have space for.

Start with the linked forum report and trace the interaction between Ratekeeper and transaction logs when TLog disk-queue extension fails. Reproduce or diagnose the low-space condition and determine how the system should prevent commits without sufficient log space; done means the failure no longer permits the database to become permanently unavailable.

Written by the indexing model from the issue text.