Main endpoint for importing batch of events. The import of events is non-transactional and each event
individually may be successful or fail, ie. the process may end up only partially successful.

This API attempts to return all errors related to particular failed events in the JSON response. The errors
may be retrieved and linked to the particular event via either:

• its 'eventId` (id of the event assigned by the client, it is optional but strongly recommended, see below)
• its relative index in the batch (automatically set up by the API)

Depending whether the eventId field is filled for each event, following two approaches can be used in
order to fix and import failed events:

• (recommended) if eventId is filled, the client may take the whole batch as was sent, fix the
  failed events as needed and send the whole (fixed) batch again. The API guarantees that only the
  last version of the particular event with particular eventId will be used (failed or not), ie. the event
  may be sent multiple times, only last version is considered valid and only the last one will be used.
• if the eventId is not filled, the client must review the failed events, fix them and send them
  again in a new batch (without the events that succeeded), possibly repeating this process until all events
  are sent. Client must not send events that were successfully imported multiple times if eventId is not set,
  because API cannot differentiate them and would import them multiple times potentially compromising the
  learning process of final recommendation model.

Due to the non-transactional and incremental nature of this endpoint, this endpoint always tries to return
HTTP 200 code even if all events sent in the batch actually failed. Other HTTP error codes are returned
only in case of hard failures (such as invalid JSON, internal error, server outage etc).

If the event is not present in the list of errors in the JSON response, the event was successfully and
persistently imported.

The importId of this batch may be set arbitrarily by the client. Events in different batches that share the
importId are considered to belong into same group. There is no need to register the importId before calling
this endpoint since related logical infrastructure is created on-the-fly upon first using the chosen importId
by the system.