implement priority queues without necessarily committing to codel.

    - Add caching in RouterAddress with secondary caching
      in Addresses; use caches to speed up transport bids,
      blocklist checks, geoip lookups, and profile organizer checks
      (ticket #707)
    - Limit IP cache size in TransportImpl
    - Clear caches at shutdown
  * RouterAddress: Remove unused expiration field to save space

increase SimpleByteCache default size.

  - Centralize priority definitions
  - Raise netdb store and reply priority
GarlicMessage:
  - Add notes about GarlicMessageHandler and HandleGarlicMessageJob
    being unused in practice

  - Limit queue sizes
  - Add stat for overflow
  - Remove some stats
  - Change pumper to LinkedHashSet for efficiency
    (like NTCP Reader/Writer)
  - Limit messages pumped per cycle to increase
    round-robin fairness
  - Comment out some unused code
  - Javadoc

  - Add logging of drops
  - Set drop stat to delay of dropped item
  - Add no-drop priority

  - Override poll() and drainTo()
  - Tweak stats
  - Add PriorityBlockingQueue version

    - Move all remaining uses to SimpleTimer2
    - Deprecate

            to branch 'i2p.i2p.zzz.test' (head 096242c22aa550274cb383a6a0c984cef07ae08c)

Use for stops, starts, and updates, and others.
Mark all restarts on graphs using the event log.

                Include test case

   - Add new flood-resistant KBucket trim policy
   - Limit received MsgID size

   - Limit UDPSender queue size
   - Increase UDPSender max packet lifetime
   - Clear UDPSender queue before sending destroys to all
   - Increase PeerState queue size so large streaming windows
     don't get dropped right away, especially at slow start
   - Various improvements on iterating over pending outbound
     messages in PeerState