index naming

[lukavdplas]

This makes some changes to how indices are named, and how es_index and es_alias are used.

The main reason was that the behaviour for database corpora needed to change, e.g. to allow duplicate names for database corpora in the future. I also wanted to change es_alias to work better for use cases like parlamint or course explorer.

• es_index is now optional, and intended to customise/override the default name. For Python corpora, the name is based on the corpus name, e.g. textcavator-times, for database corpora it's based on the ID, e.g. textcavator-custom[3]. Bracket characters are to avoid overlap with corpus names or version numbers.
• es_index is no longer set when you create a corpus via the API (but you can still set it from the admin site). The generated index name is not derived from the corpus title, so this resolves Corpus form: updating corpus title orphans index #1824
• Most test corpora no longer set es_index. Incidentally, that means you can use them in development too, without deleting the index every time you run unit tests. (Because the tests use a different application prefix.)
• es_alias now specifies an additional alias, rather than replacing the alias name. I think this is closer to how we use it. This means the corpus is still searchable under its own index name, as well as the alias.
• For a corpus with es_alias, the index --prod --rollover and alias commands now only remove the alias from earlier versions of the same corpus. (Not from indices that belong to other corpora.)
• Indexing with --prod fails with a clear message if you previously created an unversioned index for the corpus (close Indexing: no catch if intended alias is already used as index name #1947)
• In production mode, you can now create a new index version when the active version is not the latest one. (close Versioned indexing when alias is not set to latest version #1159)
• es_index/es_alias are no longer visible in the API.
• In some places, added a catch if the index name matches more than one index.

As far as I know, this won't break anything for existing corpora, because they all still specify es_index.