Avoid race conditions between API use and background processes

[timbru]

Probably use a queue for commands, and some notification mechanism (polling?) to know when a command is done.

This should also help future multi-node setups. Essentially we want to be sure that commands for a specific CA are processed in order.

For the pilot release it's not a huge deal if the occasional command fails because it's trying to update an out-of-date CA, because the command will just be tried again.

Still, this should be solved for 1.0

[timbru]

Seems to have been caused by a double trigger of requesting entitlements when parents were added. Fixed in 0.5.0