I came across this thread after I managed to solve the same problem without a library (well not a .gz specific one any rate):
Private wrk_jo As JavaObject
wrk_jo = wrk_jo.InitializeNewInstance("java.util.zip.GZIPInputStream", Array(wrk_AWS_S3_B4J.Payload_inputstream))
Private wrk_out As OutputStream
wrk_out.InitializeToBytesArray(0)
File.Copy2(wrk_jo, wrk_out)
Private wrk_bc As ByteConverter
Gen_log_file_records = wrk_bc.StringFromBytes(wrk_out.ToBytesArray, "UTF-8")
wrk_AWS_S3_B4J.Payload_inputstream is an input stream - in my case via a http job to read an AWS Cloudfront log file (which are always supplied as .gz) from an AWS S3 bucket.
Gen_log_file_records is the extracted records in string form, each record separated by a CRLF (Chr(10)).
As noted in my comments I got this by asking Windows Copilot - took 2 or 3 iterations but it knows B4J to some extent - although initially it did dream up a method that does not exist. When I pointed this out it thanked me and came up with the above.
I continue to be impressed with it.