client shouldn't log 504 (timeout) errors when calling watch on the server api.

[tealeg]

The watch function will generally timeout if nothing changes for a particular prefix (by default this will happen after 30 seconds). The client code doesn't handle the 504 explicitly (it does handle timeouts in its own context), and thus it ends up logging all the timeouts, which is annoying and misleading.

We should make it simply ignore those timeouts, or perhaps just log them at DEUG level.