(low-priority) Improve warning handling for when taking log(0)

When running idq-batch, I ran into these warnings that Reed explained could be ignored:

/usr/local/lib/python2.7/dist-packages/idq/calibration.py:455: RuntimeWarning: divide by zero encountered in log
  samples = np.log(beta.rvs(ag, bg, size=num_mcmc)) - np.log(beta.rvs(ac, bc, size=num_mcmc)) + dscale
/usr/local/lib/python2.7/dist-packages/idq/calibration.py:455: RuntimeWarning: invalid value encountered in subtract
  samples = np.log(beta.rvs(ag, bg, size=num_mcmc)) - np.log(beta.rvs(ac, bc, size=num_mcmc)) + dscale