according to my local benchmarks, it doesn't affect throughtput anyhow, but leaves constant 1 allocation/op in decompressor benchmarks

this conversion doesn't do anything, and removing it doesn't fail any of the tests

as io.Copy is just an alias for io.CopyBuffer, but passing nil instead of the actual buffer, it is being allocated internally. So better to pass it manually

this operation does nothing, so remoed it

previous variant wasn't actually adding a big misaccuracy, but better to place it anyway just before a benchmarking code

instead of returning func ptr, calling the method directly. This also doesn't bring a lot of performance gains, but on microbenchmarks may be visible

this improves readability of the code. Also added a debug print in case next step enumeric is unrecognized. However, everything will silently fail even without this print

…nually

…unks array back

as flate reader implements WriterTo interface, there's completely no difference what to use