[REVIEW] 16510 mdb incorrectly prints the IPv6 Traffic Class

[REVIEW] 16510 mdb incorrectly prints the IPv6 Traffic Class

[Bill Sommerfeld]

Issue: https://www.illumos.org/issues/16510
CR: https://code.illumos.org/c/illumos-gate/+/3450
Diff: 
https://code.illumos.org/~diff/20990cf4643ec3bb8a8e09fc384ec3459b095b5e

This is a simple fix to how mdb prints the traffic-class field of the
IPv6 header.  This is more or less the same bug as 16508 in a
different part of the code.

For background, the traffic class field is 8 bits wide but is not
byte-aligned - it's split between the low order 4 bits of the first
byte, and the high order 4 bytes of the second byte.

Our ip6_t starts with a uint32_t ip6_vcf (for "version, class, and
flow id") and there are endian-dependant mask constants to extract the
bits of interest.

The original code was doing operations in the wrong order - masking,
then shifting, then byte-swapping; on a little-endian host this always
results in zero.

With a network-order mask, the correct sequence is mask, byteswap, then 
shift.