Improves ipv6 header parsing

[nbrownus]

• Handle both types of header lengths standard ipv6 and IPSec AH
• Deal with fragments in the same way as the ipv4 path (first fragment knows the protocol, 2nd fragment gets marked as a fragment)
• Use fixed error types to avoid the heap allocation on error
• Improves test coverage for parseV6

Notably this also removes the header processing limit. The prior limit of 24 was an arbitrary choice and I was unable to determine a defensible limit. Worst case scenario we walk the total size of the packet 8 bytes at a time.

I included a benchmark to show how long a typical packet will take to process and how long a packet with 200 HopByHop headers would take.