[meshnet] Manage system-level kernel and link tunables for packet performance

[kraney]

Note: this PR stacks on top of #733 . Please merge that one first (which should make the shared commits disappear from this PR.) The incremental change of this PR starts with sha:b3c915457042d606dd0dd9bb659ed1ea619af81e.

    Implement kernel sysctl tuning in meshnetd for high-density, high-throughput network meshes.

    Reason for Change:
    - Default Linux link queue lengths, netdev backlogs, multicast membership limits, and
      strict rp_filter settings cause packet drops during netdev creation bursts and multi-link tests.
    - IPv6 DAD (Duplicate Address Detection) and Router Solicitations introduce 1-2s readiness
      delays per netdev when launching virtual topologies.

    Key Changes:
    - System & Kernel Tuning (`utils/wireutil/sys_tune.go`):
      - Sets `RLIMIT_NOFILE` to 1,048,576 open file descriptors.
      - Sets `netdev_max_backlog` to 10,000 for device queue bursts.
      - Sets OS send/recv max & default buffers (`rmem_max`, `wmem_max`, `rmem_default`, `wmem_default`) to 16 MB.
      - Sets ARP/neighbor GC thresholds (`gc_thresh1`=1024, `gc_thresh2`=4096, `gc_thresh3`=8192).
      - Sets `igmp_max_memberships` to 10,000 and `mld_max_msf` to 4,096 to prevent multicast group netdev drops.
      - Sets `rp_filter` to 2 (Loose mode) to allow asymmetric routing across multi-link meshes.
      - Disables IPv6 DAD (`accept_dad`=0) and RS floods (`router_solicitations`=0) for instant link readiness.
      - Sets `ipv6/route/max_size` to 1,048,576 for large routing tables.
      - Configured link `txqueuelen` (default 10,000) across TAP, vEth, and vxLAN interface creation.
      - All tunables are overridable via environment variables without recompilation.

[kraney]

FWIW, I created a throughput test comparing before & after for grpcwire. ("Before" being before #731 , "after" being after #734 , including the full set of improvements.)

Before:

After:

[bstoll]

Can we move all of this to documentation instead (explaining why we are changing these things?) I think these kind of settings are really up to the user to implement - we should not be forcing tuning settings in code like this. We could make changes to cloudbuild/internal.pkr.hcl and cloudbuild/external.pkr.hcl if there are changes that will improve performance measurably in the VM images.

I don't think many of these changes are relevant or helping: ARP/ND cache GC thresh doesn't do much, rp_filter is not a performance knob, ipv6/route/max_size should not be changed, default/router_solicitations , accept_dad, mld_max_msf, and igmp_max_memberships don't do much in the host OS and are not going to change performance. RLIMIT_NOFILE seems like its changing something for a limit we've never hit?

The txqueuelen, netdev_max_backlog, and [wr]mem_(default|max) I can imagine are going to affect performance. I think setting the initial window size to 16MB is a little excessive, but increasing the max seems OK.

[kraney]

I may have gotten overzealous trying to clear the last packet drops. Let me revisit