Section 4: Clarify statement about one lexical entry per stem variant

[jmccrae]

From 1st public review (Matteo Pelligrini):

"In this definition, we assume that we have one lexical entry for each stem variant so that an inflection rule whose base type doesn't match will not fire if its lexical entry doesn't fire."

This is unclear — why would we have one lexical entry for each stem variant? Needs clarification/rewording.