Empty feature vector is extracted as 1 observation

[racinmat]

	j1 = JSON.parse("""{"a": 4}""")
	j2 = JSON.parse("""{"a": { "a": "hello", "b":[5,6]}}""")
	j3 = JSON.parse("""{"a": [1, 2, 3 , 4]}""")

	sch = schema([j1,j2,j3])
	ext = suggestextractor(sch)
	a = ext(j1)
	nobs(a[:e3].data)

nobs(a[:e3].data) should be 0, because it should be missing, but currently it's 1. This should be fixed by proper implementation of missing values

[racinmat]

This is expected behavior, because observation with missing value is still observation.