Improve (re)validation of nodes

[kdeme]

Currently nodes are (re)validated by a periodic ping and removed from the routing table at first failure to reply.
This should be improved, e.g.:

1. No action should be taken on first failure. It should be marked stale after n (=5 ?) failures.
2. The replacement cache should be used: When a node goes stale, it should be replaced with a (non stale?) entry from the replacement cache, if there is one.
3. If there is no entry in the replacement cache, no immediate removal of the node should happen. Instead it should be removed when a new node for that bucket is added.

Considering that this replacement system (and kbucket ordering in general) is not really used currently (nodes only get added to it) and that there are no tests, it is possible that there are some hidden issues in the current code.
There also seems to be no current limitation on adding to the replacement cache.

Also the nodeToRevalidate should probably target least recently updated bucket.

[kdeme]

Use of replacement cache is in place, and replacement cache is limited to 8 nodes.

nodeToRevalidate targeting least recently updated bucket has not shown to give any improvement on current testnets.
More tests are in place now.

see #247 and #261

1. and 3. are not currently done. 3. specifically might be an issue when the own connection drops or is just terrible.

Issue for those is created here: #262