Test: lrmd: bump regression test's command timeout yet again

The issue appears to be the Travis CI infrastructure; I can't find anything
intrinsically slow in the code. Longer-term, it would probably be better to
tailor timeouts to specific commands or resource classes.