Benchmark mtl code run via transformers and effectful

[arybczak]

Turns out that running unspecialized code written with mtl style effects in mind via effectful's Eff is always faster than running via a transformer stack. Crazy.

countdown.1000:

mtl (transformers)
  shallow:        OK
    51.1 μs ± 2.9 μs, 477 KB allocated,  92 B  copied, 6.0 MB peak memory
  deep:           OK
    409  μs ±  33 μs, 3.0 MB allocated, 870 B  copied, 6.0 MB peak memory
mtl (effectful)
  shallow:        OK
    36.4 μs ± 2.9 μs, 260 KB allocated,  71 B  copied, 6.0 MB peak memory
  deep:           OK
    36.9 μs ± 2.9 μs, 267 KB allocated, 179 B  copied, 6.0 MB peak memory

filesize.1000:

mtl (transformers)
  shallow: OK
    973  μs ±  87 μs, 7.2 MB allocated, 349 KB copied, 7.0 MB peak memory
  deep:    OK
    2.20 ms ± 175 μs,  15 MB allocated, 854 KB copied, 8.0 MB peak memory
mtl (effectful)
  shallow: OK
    818  μs ±  44 μs, 6.5 MB allocated, 196 KB copied, 8.0 MB peak memory
  deep:    OK
    883  μs ±  77 μs, 6.6 MB allocated, 210 KB copied, 8.0 MB peak memory

[Kleidukos]

@arybczak I'll have to write something about how the new MTL instances on Eff allow for progressive incorporation of effectful in a codebase.