reindexing large PopIt instances exhausts Elasticsearch's memory

[mhl]

To quote @struan:

[T]he re-indexing uses the elasticsearch's bulk re-index facility so it's throwing all the json for each collection at elasticsearch in one go. We did this as re-indexing each document individually caused everything to crash to a halt. It may be that we need to do split up the collections a bit as a first stage and then re-index those bits.

Related:

• https://github.com/mysociety/Non-Site-Specific/issues/551
• https://github.com/mysociety/popit/issues/781