Add WARC compressed record length to the extraction

[anjackson]

Can we use e.g. a counting stream-reader to work out how long each WARC record is (compressed?).

[tokee]

We (the Royal Danish Library) would like this for CDX API support. It is definitely possible, and I am 75% sure the functionality is already there, just buried at an unknown level in the convoluted stack of IndexStreams that is used. I'll see if I can find the time to dig into this.