Proposed amendment to <certainty> values

[Art Botterell <acb@incident.com>]

The current values for the <certainty> element are accurate enough, 
but they provide no mechanism for distinguishing actually observed 
events from high-confidence forecasts.  This creates a bit of 
dissonance, e.g., when a report of an actual earthquake that just 
occurred is described as "Very Likely."  So I propose that the values 
be changed to the following:

	Observed - Reported as actually having occurred by a credible source.
	Likely - Probability > ~50%
	Possible - Probability < = ~50%
	Unlikely - Probability ~ 0
	Unknown - Probability unknown

This would have the effect of merging the existing "Very Likely" and 
"Likely" categories, thus also getting rid of that single two-word 
label.

Alternatively, the "Observed" value could simply be added to the 
existing scheme.

- Art