libxl-migration-fdset-staging

libxl: Enable save/restore/migration of a restricted QEMU + libxl__ev_qmp_*

The real meat in this patch series start with patch
"libxl_qmp_ev: Introduce libxl__ev_qmp_start() to connect to QMP"
which implement libxl__ev_qmp_* functions to turn the QMP client into
asynchronous mode.

This comes with two examples on how to use it:
* "libxl_disk: Have libxl_cdrom_insert use libxl__ev_qmp"
  with patches:
  - "libxl_qmp: Implement libxl__qmp_insert_cdrom_ev"
  - "libxl_disk: Cut libxl_cdrom_insert into step"
* "libxl: QEMU startup sync based on QMP"
  which can use QMP to find out when QEMU as started.
  this requires: "libxl_dm: Pre-open QMP socket for QEMU"
  But that only works with dm_restrict=1 as explain in the patch.

The first few patches do some cleanup and fixes of the current qmp client
implementation, mostly because it bothered me as I think we should remove the
current implementation. There is also two patches to allow to save a restricted
QEMU, but that would need to be converted over to libxl__ev_qmp_*.

There is still one bug that I haven't fix yet. When creating a guest with
dm_restrict=1, the call to libxl__qmp_initializations() is going to fail
because libxl is still connected to the QMP socket. But libxl doesn't care
about failure, and that just mean that `xl console` will not work and vnc will
not have any password. save/restore of the same guest will works fine because
libxl__ev_qmp_* will have an oportunity to disconnect from the socket before
libxl__qmp_initializations() is called.

Cheers,