Store sentence IDs in the correct named entity field

Prior to this change, when named entities were created, their appearing
sentence wasn’t being associated because we were trying to store its ID
in `sentenceId` instead of `sentence_id`, the latter of which is the
actual Postgres column name.

However, by explictly snake-casing at this point, we may be breaking
some of the correct stitching of snake-cased Postgres fields to
camelCased object attributes, and so we need to debug whether this is
all associated together correctly by Sequelize and queryable by GraphQL.

Affects #49

src/server/workers/openedCaptions/index.js

@@ -81,7 +81,7 @@ async function processSentence(content) {
   ents.map(ent => models.NamedEntity.create({
     entity: sentence.substring(ent.start, ent.end),
     type: ent.label,
-    sentenceId: storedSentence.id,
+    sentence_id: storedSentence.id,
     model: 'en_core_web_lg',
   }))
 }