This was causing some (most) d710s to load images from boss rather than
the subboss.

One difference: we don't make all the symlinks to script_wrapper.py (like
node_avail et al).

Also, this is going to require m2crypto -- the right version -- on client
images.  If the scripts don't work on many images for now because
m2crypto is not there, fine for awhile... but we should move in this
direction.

We now make a pass through the implemented_by tree setting port speeds
that the user left up to assign (i.e., that are 0).  We start at the
lowest level links and assign default speeds.  Then the speed values for
the start and end segments for each path get propagated to the link
being implemented by the path.  Should be pretty generic for any number
of layers.  We warn if the higher level links don't support those speed
settings (if they don't, we're gonna have a problem; but I think assign
protects this from happening).

Backward compat fall back when no frisbee upload...FAIL.
Upgrade MFS so no backward compat needed...FAIL.

Add new -u and -U options to include (or exclude) nodes that have
a given feature.

Nice excuse to learn MySQL subqueries.

Now that we have logical wires, we *must* actually make sure that the
VLANs created on the per-experiment switch are removed at swapout (i.e.,
are at least removed from the vlans table), so that any logical wires can
be safely removed.  For now, we actually invoke snmpit to remove them.
This is probably a bad idea, because who knows what state the user left
the switch in prior to swapout.

The bugfixes are just me forgetting that \ does not continue a line in
perl :).

1. Support for PUT.

The big change is support for uploading via the master server, based heavily
on the prototype that Grant did. Currently only host-based (IP-based)
authentication is done as is the case with download. Grant's SSL-based
authentication code is "integrated" but has not even been compiled in.

The PUT protocol allows for assorted gewgaws, like specifying a maximum size,
setting a timeout value, returning size and signature info, etc.

There is a new, awkwardly-named client utility "frisupload" which, like the
download client, takes an "image ID" as an argument and requests to upload
(PUT) that image via the master server. As with download, the image ID can
be either of the form "<pid>/<emulab-image-name>", to upload/update an actual
Emulab image or it can start with a "/" in which case it is considered to
be a pathname on the server.

On the server side, the master server takes PUT requests, verifies permission
to upload the image, fires up a separate instance of an upload daemon (with
the even catchier moniker "frisuploadd"), and returns the unicast addr/port
info to the client which then begins the upload. The master server also acts
as a traffic cop to make sure that downloads and uploads (or uploads and
uploads) don't overlap.

This has been integrated into the Emulab "create image" process in a
backward-compatible way (i.e., so old admin MFSes will continue to work).
Boy, was that fun. One not-so-desirable effect of this integration is that
images now traverse our network twice, once to upload from node to boss and
once for boss to write out the image file across NFS to ops. This is not
really something that should be "fixed" in frisbee, it is only "undesirable"
because we have a crappy NFS server.

What has NOT been done includes: support of hierarchical PUT operations
(we don't need it for either the elabinelab or subboss case), support for
uploading standard images stored on boss (we really want something better
than host-based authentication here), and the aforementioned support of
SSL-based authentication.

2. Other tidbits that got mixed in with PUT support:

Added two new site variables:
    images/frisbee/maxrate_std
    images/frisbee/maxrate_usr
which replace the hardwired (in mfrisbeed and frisbeelauncher before that)
bandwidth limits for image download. mfrisbeed reads these (and the
images/create/* variables) when it starts up or receives a HUP signal.
These could be read from the DB on every GET/PUT, but they really don't change
much and I needed something to test the reread-the-config-on-a-HUP code!

Fixed avoidance of "problematic multicast addresses" so it would actually
work as intended.

Lots of internal "refactoring" to make up for things I did wrong the first
time and to give the general impression that "Wow, Mike did a LOT!"

This is specifically intended for elabinelab right now where configurations
with an "inner control network" need a multicast router for inner frisbee
to work. There is (an undocumented) NEEDMROUTED defs variable that controls
whether the machinery is triggered or not. It is not documented because, though
in theory this would work for a real boss setup, it has not been tested and
each install would probably need a custom mrouted.conf file.

For an inner elab, this variable is automatically set in defs-elabinelab
if there is an inner control network. The result is that the mrouted port
will be installed, it will be enabled in rc.conf, and a stub mrouted.conf
file is created (to force mrouted to ignore the real control network).

translation table.

Per-experiment switch stacks only come into being if the experiment
actually has a switch allocated to it.  If not, tbswap and snmpit
should function unchanged.  If there is a per-experiment stack that needs
configuration, we first invoke normal snmpit in the normal place, but we
use the new snmpit option `--skip-supplied' in combination with -S to skip
the per-experiment stack.  We then configure the per-experiment stack by
itself with -S after os_setup has completed.

There are some new functions in the db backend stuff to create, modify,
and remove per-experiment switches.

There is some new code in snmpit to do the --skip-supplied filtering.  I
also put all the -S, -i, and --skip-supplied stuff into portstats...
because we also can't call portstats on a per-experiment switch in tbswap;
otherwise it will hang and/or fail.

TODO: Track down floating point conversion issue

URNs in component_ids should now always refer to the physical node rather than the virtual container.

showstats page way way faster.