Handle missing values in content field

If the content field is missing in some documents, the script will fail during collection. This should be handled gracefully (i.e. treat the value as an empty string).

If the value is missing from *all* documents, training word2vec models will fail, obviously, but this is not communicated clearly in the log. Maybe an extra check would be useful.

Start at the script's Elasticsearch collection path and follow how the content field is read before training begins. Handle missing content as an empty string, and add a clear log message when no document contains content; verify both cases through the existing script workflow.

Written by the indexing model from the issue text.