Bridges are still tried to measured before their descriptors arrive

At #138 (closed) we reduced the number of failed measurements by queueing the bridges to measure when their descriptor arrives, but it still queues bridges in the database when there aren't enough bridges in that queue, resulting in many measurements as "failed" because of "No such router", which also produces tpo/anti-censorship/rdsys!137 (closed).

We could try to solve this in different ways:

• returning functional=true when the measurement error is "No such router"
• the same, but only when we have a previous measurement, not too "old" without that error
• try to set all known bridges to tor configuration, so that it gets the bridge descriptors before it tries to measure the bridges (not sure this'll work)