Separate archiver/compressor auto-detection behavior

[thrau]

The commons stream factories provide methods for creating input-streams based on auto-detection. This functionality is currently used without any kind of consideration. Archivers can extract anything that is auto-detectable by the ArchiveStreamFactory, because CommonsArchiver creates the streams by passing the InputStream of the file to extract, rather than the archive format.
While this may be a convenient behavior in some cases, its confusing, unclean and can cause issues where archive/compression types may actually not be auto-detectable (all the new compression formats, or JAR for example).

ArchiveStreamFactory cannot tell JAR archives from ZIP archives and will not auto-detect JARs.

Auto-detection should but be segregated and properly documented.