consider a linearly scannable archive alternative to kzip

[jaysachs]

Ingested data is already large, and only getting larger. Being able to process (an alternative to) kzips via a linear scan would significantly reduce I/O costs (both storage and throughput)

Requirements:

• No new compression or archive scheme; "expandable" / readable via standard archive tools
• The compilation unit data must precede all other data (which currently only includes file contents)
• The format should be reasonably compressed

Undecided at present if we'll re-purpose the kzip tool for this, or if we'll create a new tool.