* Switch from VOEventLib to voevent-parse due to significant
  string-bytes issues in VOEventLib
* Remove bad exception handling in events API for creating a
  VOEvent
* Add parameter checking to events API for creating a VOEvent
* Combine VOEvent building code for events and superevents into
  a single instance which handles both cases

Remove usage of .keys(), .has_key(), and .iteritems(), and
account for the fact that .values() produces an iterator in Python 3.

we need to dance between str (unicode) and bytes to talk to hashlib and StringIO at the same time

Add a pipeline_type field to the Pipeline model. This will facilitate
grouping of pipelines into different classes.

There are several fields that the user provides to create a
VOEvent through the API, which are used in generating the VOEvent
file, but are not stored in the database. Let's fix this!!

GroupObjectPermission and data directory cleanup should only happen
*after* the event or superevent row is removed from the relevant
table.

A web view has been created which allows EM advocates to disable
or enable pipeline submissions. This is to prevent misbehaving
pipelines from submitting bad information. The mechanism works by
preventing events from being submitted to a given pipeline, not
by revoking certificates or removing user account permissions.

New model functionality, better forms and views.

Some additional code cleanup, as well.

All calls to event.graceid() -> event.graceid.

Enforce label protection through the API for applying labels,
removing labels, and creating events and superevents with labels
attached.  We also don't allow users to reapply a signoff request
label when a signoff status label is already applied (e.g., can't
apply ADVREQ when ADVNO already exists).

Add a boolean column called 'protected' to the Label model.
Protected labels will not be allowed to be directly applied, but
will be applied as a part of another process (like signoffs).
We also create a data migration which sets some existing labels
as protected (ADV(OK|NO), (H1|L1|V1)(OK|NO))

Added a 'get_subclass' method to the Event model which is used to
get the CoincInspiralEvent, GrbEvent, etc. corresponding to that
Event.

Changing signoff form and expose/hide button on the superevent
web display to make an AJAX request to the relevant API endpoints.
This way, we don't have to maintain separate permissions and views
for web resources.

The "full_filename" property for Log objects is now called the
"versioned_filename" to improve clarity.

Adding utils to Superevent and Event models for comparing
categories and deciding if an Event is the correct category to be
part of a superevent.

Models which inherit from AutoIncrementModel can now specify
more than one constraint.  Needed for superevent categories.

Added some utilities for handling signoffs for superevents,
including alerts, labels, logs, etc.

Allow EMObservation fields ra, dec, raWidth, and decWidth to be
blank in forms, resulting in a null database entry.

If a tag's name has some "uncommon" characters in it, like
parentheses (or probably others), it will cause errors for the
javascript which renders the event pages, since it tries to use
the tag name in the element id for the delete button.  We now require
that tags only use 0-9a-zA-Z_\- in their names. May have to go through
and delete some tags which use them.

generalizing and abstracting several models in the events app, as well as fixing many field parameters

Need to make sure values are actually numeric types before
just passing them to isnan(), since it can't handle them, apparently.

gstlal wants to set eff_distance to nan when they don't have a
meaningful value for it.  Unfortunately, we can't save that in the
database because MySQL numeric types can't deal with nans. So we
check for nans in this field and set it to NULL if it is nan,
before saving it in the database.