Remove NodeNorm's code for determining the preferred name for a clique

[gaurav]

NodeNorm has always calculated the preferred name for a clique itself, even though we need to calculate this in Babel so we can use it in NameRes:

NodeNormalization/node_normalizer/normalizer.py

Lines 707 to 736 in 13ba01e

	# As per https://github.com/TranslatorSRI/Babel/issues/158, we select the first label from any
	# identifier _except_ where one of the types is in preferred_name_boost_prefixes, in which case
	# we prefer the prefixes listed there.
	labels = list(filter(lambda x: len(x) > 0, [eid['l'] for eid in eids if 'l' in eid]))
	# Note that types[canonical_id] goes from most specific to least specific, so we
	# need to reverse it in order to apply preferred_name_boost_prefixes for the most
	# specific type.
	for typ in types[canonical_id][::-1]:
	if typ in config['preferred_name_boost_prefixes']:
	# This is the most specific matching type, so we use this.
	labels = map(lambda identifier: identifier.get('l', ''),
	sort_identifiers_with_boosted_prefixes(
	eids,
	config['preferred_name_boost_prefixes'][typ]
	))
	break
	# Filter out unsuitable labels.
	labels = [l for l in labels if
	l and # Ignore blank or empty names.
	not l.startswith('CHEMBL') # Some CHEMBL names are just the identifier again.
	]
	# Note that the id will be from the equivalent ids, not the canonical_id. This is to handle conflation
	if len(labels) > 0:
	node = {"id": {"identifier": eids[0]['i'], "label": labels[0]}}
	else:
	# Sometimes, nothing has a label :(
	node = {"id": {"identifier": eids[0]['i']}}

This means we need to manually keep this in sync with Babel.

As we redesign NodeNorm's database to add clique-level properties (#302), we should see if we can use the preferred name from the compendium files instead of having to recalculate them here.