freebsd-mach-kmod-syscall-slots-spike: 10-slot lkmnosys cap audit plan

freebsd-launchd-mach has wired 10 Mach syscalls into FreeBSD's
reserved lkmnosys slot range (210-219). An 11th fails with ENFILE.
Forward-demand estimate puts total Mach syscalls at 30-40 by the
time launchd+configd+notifyd+asl are ported.

This is an audit plan, not a design decision. Outlines:

  - The 10-syscall inventory currently shipped.
  - Why the FreeBSD limit exists (compile-time-sized sysent table,
    convention not technical constraint).
  - What Apple does differently (separate mach_trap_table with
    negative syscall numbers, ~80 entries today).
  - Risk profile per expansion path: RESERVED-slot claiming
    (low-risk on low-numbered RESERVED slots, ~48 available),
    multiplexer dispatcher (one slot, internal table), base-kernel
    patch (disqualified by project rules).
  - Forward-demand estimate broken down by phase.
  - Four sub-agent audits to run before picking a path:
    A) FreeBSD syscall-extension mechanisms
    B) Survey existing Mach-syscall users
    C) XNU's own Mach-trap dispatch
    D) Impact on existing wired syscalls
  - Decision rubric with weighted criteria.

Phase G2b's routing workaround stays in place while we audit;
migration commit lands before Phase G2c.

Author: pkgdemon