suggested change to spec in light of PyTREX bug

I just fixed a bug in PyTREX pointed out by James Clark.

It came about because of a misinterpretation of the spec on my part. The
spec reads:

	M [[<t:attribute> name-class pattern </t:attribute>]] (a, c, e) if
and only if 
		c is empty, and 
		a consists of an attribute <t:n,v>, and 
		C [[name-class]](n), and 
		M [[pattern]]({}, v, e) 

I took this to mean that a could consist of any number of attributes of
which one needed to be <t:n,v> rather than that a must consist of a single
attribute <t:n,v>

Upon reflection, it is obvious it should be the latter but I wonder if there
would be value in change it to read "a consists of a _single_ attribute
<t:n,v>".

James