Fix Cisco device canonicalization regular expression on 1.9

The behaviour of the \b Regexp match operator, or rather, the classification
of / as an end of word character, changed between Ruby 1.8 and 1.9.

Most uses in our code are safe, and this will lead to equally correct
behaviour, but the Cisco device interface canonicalization code used it where
"end of string before newline" was intended.

Moving to the \Z match, which imposes the exactly desired condition,
resolves the problem.

Signed-off-by: Daniel Pittman <daniel@puppetlabs.com>