Analyse qualifier proliferation in benchmarks

[facundominguez]

Since #2488 we've noticed that several qualifiers are duplicated in fixpoint queries, and the amount of qualifiers have grown in that PR. We identified growth in the following files:

tests/benchmarks/vector-algorithms-0.5.4.2/Data/Vector/Algorithms/Merge.hs
tests/benchmarks/vector-algorithms-0.5.4.2/Data/Vector/Algorithms/AmericanFlag.hs
tests/benchmarks/bytestring-0.9.2.1/Data/ByteString/Fusion.hs
tests/benchmarks/bytestring-0.9.2.1/Data/ByteString/Fusion/T.hs

We have collected the fq files before and after #2488.

Here are some examples of duplicated qualifiers from before #2488.

Here's a summary of the growth that we have observed due to #2488, which I think is due to fac8f57 feeding more binders to the generation of qualifiers.

This ticket is to improve the generation of qualifiers so the duplication is reduced, and to analyse the options to identify and drop irrelevant qualifiers.