WW should also zap useless idInfos in the `certainlyWillInline` case.
Currently inside tryWW
we generate a new ID like this:
new_fn_id = zapIdUsedOnceInfo (zapIdUsageEnvInfo fn_id)
-- See Note [Zapping DmdEnv after Demand Analyzer] and
-- See Note [Zapping Used Once info WorkWrap]
Which is always used as the new Id except in the case where the function is guaranteed to be inlined.
Based on the referenced notes I think we should zap the binding even if we don't perform WW.