force key encoding to binary to avoid UTF-8 issues

[aaronpk]

Somehow I ended up with keys in redis that contained invalid UTF-8 characters, which is now causing problems trying to retrieve them.

My code is currently failing with this stacktrace:

ArgumentError: invalid byte sequence in UTF-8
        from /var/ruby/vendor/bundle/ruby/2.0.0/gems/redis-namespace-1.3.2/lib/redis/namespace.rb:380:in `gsub'
        from /var/ruby/vendor/bundle/ruby/2.0.0/gems/redis-namespace-1.3.2/lib/redis/namespace.rb:380:in `rem_namespace'
        from /var/ruby/vendor/bundle/ruby/2.0.0/gems/redis-namespace-1.3.2/lib/redis/namespace.rb:376:in `block in rem_namespace'
        from /var/ruby/vendor/bundle/ruby/2.0.0/gems/redis-namespace-1.3.2/lib/redis/namespace.rb:376:in `map'
        from /var/ruby/vendor/bundle/ruby/2.0.0/gems/redis-namespace-1.3.2/lib/redis/namespace.rb:376:in `rem_namespace'
        from /var/ruby/vendor/bundle/ruby/2.0.0/gems/redis-namespace-1.3.2/lib/redis/namespace.rb:338:in `method_missing'
        from /var/ruby/vendor/bundle/ruby/2.0.0/gems/redis-namespace-1.3.2/lib/redis/namespace.rb:227:in `keys'

pointing to the source of the problem being the .sub method of the key name.

I was able to get around this issue by forcing the string encoding to binary, which still successfully removes the namespace prefix from the keys.

[aaronpk]

Okay well it looks like my fix fails in ruby 1.9.3 and below for various reasons. Any better suggestions?

[aaronpk]

Okay I fixed this for all versions except 1.8.7 which doesn't have the concept of string encodings, so doesn't need fixing. I added a test case demonstrating how it failed in the first place.

[iloveitaly]

@aaronpk I know this is an old once, but could you rebase on master so the latest tests run?