signal: implement signalfd(2)

[gburd]

What

signalfd(2) was stubbed to return ENOSYS, so event loops that consume
signals through a file descriptor (a common pattern with epoll) could not run.

How

Implemented as a pollable special_file (modeled on eventfd) backed by a
global registry of live signalfd objects. OSv is a single process, so a
signalfd watches a set of signals for the whole image. kill() now checks the
registry first: if any signalfd is watching the raised signal, the signal is
queued to those fds (as a signalfd_siginfo) and their pollers are woken, and
the signal is consumed -- the default action and any handler are skipped. This
matches Linux, where a signalfd-consumed signal must be blocked and is dequeued
through the fd.

• signalfd(-1, mask, flags) creates a new fd; signalfd(fd, mask, flags)
  updates an existing one's mask. SFD_CLOEXEC and SFD_NONBLOCK are honored.
• read() returns as many queued signalfd_siginfo structures as fit in the
  buffer, blocking (or EAGAIN when non-blocking) when the queue is empty, and
  rejecting a buffer smaller than one siginfo with EINVAL.
• poll() reports POLLIN when signals are queued.

The delivery hook (osv_signalfd_deliver) is declared in <osv/signal.hh> and
called from kill() in libc/signal.cc; the stub there is removed.

Testing

tests/tst-signalfd.cc covers delivery of a watched signal to the fd (instead
of the default poweroff action), the EAGAIN/poll/queue-order paths, and the
EINVAL cases. Passes on OSv under KVM (1 and 2 vCPUs). tst-sigwait and
tst-sigaction continue to pass, so normal signal delivery is unaffected for
signals no signalfd is watching.

Note

OSv's signal model is thread-directed and minimal; this integrates at the
process-wide kill() delivery point, which is the right granularity for a
single-address-space unikernel. Signals raised synchronously in-thread (e.g.
faults) are a separate path and not routed to signalfd here.

[gburd]

Superseded by #1439 (rebased onto current master; GitHub auto-closed this one when the branch was updated). Please review #1439 instead.