Tool for choosing when to fire gulpers based on an HMM from which the
cluster server provides "observations".  Objects:

dtClusterHMM --> The states, transition table, and table of
	     conditional probabilities (probability of being in state
	     x based on observation y) will be configuration options.
	     Most of the work is done by the HMM class.

HMM --> Relatively generic tool for keeping track of the state in an
    HMM.  State is defined externally, as are the transition and
    conditional probability tables on initialization.  On update at
    time t, an observation is fed in, and this module updates the
    probability of being in state s_i according to:

    p_t(s_i) = p(s_i|obs)*(sum_j p(s_i|s_j)*p_{t-1}(s_j))

HMMLogger --> Tool for logging the observation and state in binary format
	  at each update.


CONFIGURATION FILES:
smartSampler_HMMFour --> Default dt config file: cluster HMM with four
		     states.  Contains state info and names of configuration
		     files for loading transition and probability tables.

smartSampler_HMMFour_trans --> Default state transition table for HMM.

smartSampler_HMMFour_cond --> Default conditional probability table for HMM.
