* implement stream compression

* fix zstd + pass benchmarks

* add comments and remove old API

* add kafka.CompressionCodec.Name()

* avoid a copy when the output buffer for decoding snappy compressed data is large enough

* fix xerial framing + validate against go-xerial-snappy

* remove unused function

* PR feedback

* cleanup APIs + improve benchmarks

* optimize gzip reader

* fix snappy compression codec

* support running decompression benchmarks alone