Propose errata for emma attribute

We've found that generating a "simple" XML/DOM wrapper of the results for
the SpeechRecognitionEvent emma attribute is more challenging to implement
than originally thought. I propose that the UA may return undefined for
implementations in which the speech recognition engine does not supply
emma.  In this case, emma doesn't provide any additional information than
is already available via the API in the results attribute.

Specifically, I propose adding the phrase: ", or if the recognizer does not
supply EMMA then the user agent may return undefined."

Here it is in context [1] ...

emma attribute
EMMA 1.0 representation of this result. [EMMA] The contents of this result
could vary across user agents and recognition engines, but all
implementations must expose a valid XML document complete with EMMA
namespace, or if the recognizer does not supply EMMA then the user agent
may return undefined. User agent implementations for recognizers that
supply EMMA must contain all annotations and content generated by the
recognition resources utilized for recognition, except where infeasible due
to conflicting attributes. The user agent may add additional annotations to
provide a richer result for the developer.

[1] https://dvcs.w3.org/hg/speech-api/raw-file/tip/speechapi.html#dfn-emma