This is an experimental change for investigating the performance
consequences of using more strictness in UniqFM.
On my machine I've already been surprised by the following stat failuresWhile not making a performance difference at O2 using foldl'
(stat not good enough):
TEST="T12227 T12545 T3064 T5030 T5321Fun T5631 T9872a T9872b T9872c"makes it clear that these are strict.