Cross datacenter setups are possible with Redis-cluster, but there aren't that many well documented setups that people are using in production, nor a standard approach people agree with across the board.
I'm pretty sure it should be possible to do what you're asking, but I'm not sure how well it will play with cluster features, such as electing new masters when one goes down. The approach I'd first look at, would be to just add a handful of slave nodes to the second data center and see if you can find a way to configure those nodes to never be promoted to masters. That would require at least 1 slave node per write node.
You couldn't be able to get away with less nodes in the second datacenter. Though you could still put multiple nodes on a single box (run multiple redis instances on different ports). That might let you still use X dedicated nodes in DC1, and X slave nodes in DC2, but with 2 nodes per box in DC2, so only X/2 hosts used.
Another option might be to look at Dynomite.
http://www.dynomitedb.com/That's a very different approach from using Redis-Cluster, but is built with multi-dc replication in mind. If it's a cleaner fit for your use case, that might be a productive option.