Fixes for inefficient reads of ZIP entries

[chocolatkey]

Fixes readium/cli#27

• Fixes bugged Read calls that aren't obligated to return the entirety of a resource
• Add zran-like reading of ZIP entries. This only applies to the start offset, not the total length of the resources. Particularly useful for e.g. streaming compressed videos where a browser will send Range: bytes=0- then abort when satisfied, which is already hopeless for us to handle. At least when users scrub through video/audio, we can take a Range: bytes=XXXXXX- header and start the remote range request for the resource closer to the start of the byte range. This happens in increments of 1MB
• Add CRC32 checksum function for CompressedResource to make direct copying of compressed resources to new archives easier

Still getting ERROR failed streaming asset error="resource: error 500: flate: corrupt input before offset 5", need to fix that before this becomes a proper PR because there's a bug somewhere

[chocolatkey]

All that needs to be done is using a readium repo for gzran (which will become zran)