Fix #2626, #2623: handle socket fields sin6_len & sin_len, when present on OS

[LeeTibbert]

Two socket fields, sin6_len & sin_len which are commonly available on BSD-derived
systems, such as macOS & FreeBSD, are now handled correctly.

Two new tests InTest.scala and In6Test.scala are introduced to verify the changes.

A benefit of this now proper handling is that socket methods sendto() and recvfrom() work correctly
on those two operating systems (Issue #2623, fixed on Linux & Windows in a prior PR.)

[LeeTibbert]

Moving to Work In Progress (WIP). CI revealed an error.
I want to take my time to understand it and come up
with a proper fix. I have a hack fix in my sandbox, so I know there is at least one
solution.

[LeeTibbert]

Modulo a few spurious CI errors, this PR is ready for review.

Partest is reliably (> 50%) killing qemu. The SEGFAULT
appears to happen after a Java garbage collection, usually
with commix. Commix doesn't like memory pressure?
Use G1GC (in .sbtopts)? Specify more memory (in .sbtopts)?
Skip Partest on qemu?

I got 83 out of 85. One apparently hung linux test, and one
failing qemu arm64. Everybody else was green.

I am firing off another CI run and will check it later.
I am pretty certain the bugs are in CI and no longer mine,
(earlier CI did catch real macOS bugs in my code).

[WojciechMazur]

Looks good, I believe the issues in the CI are not related to this change