To get shark on LZO files working (I have it up and running with CDH4.4.0) you first need the hadoop-lzo jar on the classpath for shark (and spark). Hadoop-lzo seems to require its native code component, unlike Hadoop which can run non-native if it can't find native. So you'll need to add hadoop-lzo's native component to the library path too.
Here's an excerpt from my puppet module that does these things. Edit accordingly and put these two rows into your shark-env.sh