Investigate some cases for performance

[harendra-kumar]

1. aheadly applicative tests (maxBuffer?) in Prop.hs seem to run quite slowly
2. maxBuffer 1 test for aheadly seems to run very slow

We need to check if there is an opportunity for optimization in these cases.

[harendra-kumar]

Some other performance related investigations:

1. The benchmark routine "benchIO" in streaming-benchmarks as well as in streamly benchmarks needs to be INLINEd for streamly and conduit but not for vector. Why?
2. parallely/zip benchmark has degraded from 32ms to almost 50 ms over time. need to see if this can be improved again.
3. When mapM is used with unfoldrM and both are used aheadly, the results are much
   worse compared to asyncly.
4. filterSome with aheadly seems to perform pretty badly when using a 1000x1000 iterations.