sawsim/discussion.tex: Rework Bell testing to match scaffold phrasing

sawsim/discussion.tex: Rework Bell testing to match scaffold phrasing

This fills in the derivation of the Gumbel pdf from the exponential
hazard function.  In the scaffold section, we rephrased the hazard
function and then cited NIST:gumbel to make the transition to the pdf.
Maybe this derivation would be better of in an appendix?  Or dropped
entirely?  I'll leave it for now.