This allows for specifying a "GRD_MANAGER" variable in the user code that
will be used to set the initial MANAGER record, with an initial MODE of
MANAGED.

keep everything in one place

since shmems are initialized to value "0", we want to do all pre-exec
checks together before worker status moves out of INIT.  In particular, we
move the SUBNODES value checks to this section.

We then require that worker not be in INIT for ACTIVE to be True.

All guardctrl handling moved into the python module (both local and remote).

Use json encoding to pass arguments to SSH_ORIGINAL_COMMAND, which is a much
more robust way of handling argument spaces, and unbreaks various option
handling.

The guardlog command is re-purposed to exclusively handle opening logs in
an xterm.

The same-state redirect (when there is a request for the current executing
state that causes it to redirect back to the top of the state and reexecute
the main method) is somewhat unintuitive and has caused problems:

https://git.ligo.org/cds/guardian/issues/18

This change requires that an edge exist from a state to itself in order
for a same-state redirect to take place place, e.g.:

edge = [('STATE', 'STATE')]

A new test was added for this change.

Not sure why this has become a problem now, but with timeout set to zero
the connections time out immediately, which is obviously not desired.
The only affect should be slowing the test suite slightly.

unclear what changed...  worrisome

The system doesn't need to be loaded to provide this info.

Also allow printing channels even when system is not specified.

networkx 2 dropped the {nodes,edges}_iter methods, with the nodes/edges
methods now returning Views.  We can just use the non _iter methods from
earlier versions that just return lists of nodes and edges, as these
methods are available in all versions (although they return lists instead
of Views in versions earlier than 2).

arrived AND done.  closes #10.

This adds three new summary status bits:

ACTIVE: node operation is within normal operational parameters
READY: all sub nodes are reporting OK status
INTENT: request is equal to nominal state, and not stalled

These should help improve downstream system tracking, and are particularly
useful for the IFO top node reporting of overall IFO status.

SUBNODES_TOTAL and SUBNODES_NOT_OK give the total count of subnodes and
the subset that are not reporting OK status, respectively.  This info is
pulled from the manager interfaces in the worker subprocess.

These are important post-facto diagnostics

This small change just prevents this pseudo state from showing up in the
drop-down menu.  The state is still requestable, and none of associated
logic is modified.

remove the requirement of needing the path.  These screens are
basically always embedded in situations where path is not needed (or
desired).

should fix:

https://bugzilla.ligo-wa.caltech.edu/bugzilla3/show_bug.cgi?id=1115

SPM channel summaries, and re-enable NOTIFICATION

clarify int microseconds.

Here we do batch updating of the driver PVs, rather than after every setitem.
driver.updatePVs is a heavy operation, proving to be most of the CPU load
during profiling.  Batch updating reduces the number of updatePVs calls by an order of magnitude, from
roughly 100 per cycle (when coupled with cas.__setitem__) to roughly 10 in
the current implementation.

Unfortunately there's some way that updatePVs seems to affect client writes
that is not yet understood.  We should only have to call updatePVs twice in
the main loop to push out all updates.  But some weird behavior was observed,
where client writes didn't seem to take affect.  This needs to be investigated
more, but in the mean time calling updatePVs around process seems to behave
ok.

The ordering in the channel updates as seen by the tests changes slightly,
since all updates happen as a batch rather than individually when they are
set.

Was producing too much logs with no real additional benefit.

and cleanup, add FIXME

INIT behavior is different than other states, and is not intended to be set
as a fixed request state.  Throw error on attempt to set it as initial
request.