feat: Use peliasQuery analyzer for address_parts.street

This is an experimental change, but one that has good impact when paired
with pelias/schema#446.

Essentially, this change continues the trend we have started for the
`name.*` and `phrase.*` fields to generate synonyms _only_ at index
time.

It may be the case that variations on input vs data text (for example
`crt` vs `ct` vs `court`) may cause different synonyms to be generated
by the same analyzer.

Many queries, especially `match_phrase` queries, will require that _all_
of those generated synonym tokens must match. This is often not
desirable.

query/search_defaults.js

@@ -41,7 +41,7 @@ module.exports = _.merge({}, peliasQuery.defaults, {
   'address:housenumber:boost': 2,
   'address:housenumber:cutoff_frequency': 0.01,
 
-  'address:street:analyzer': 'peliasStreet',
+  'address:street:analyzer': 'peliasQuery',
   'address:street:field': 'address_parts.street',
   'address:street:boost': 5,
   'address:street:slop': 1,