Always de-reference the value got from get_mapping_path().
Found by Magali.

translation ID, and ALWAYS update it to the correct translation ID
later, after storing the translation.

Avoids problems (the canonical_translation_id being duplicated) when
transferring transcripts with translations from another database.

Amonida found this.

  Fix store() so that the canonical_translation_id in the transcript
  table is updated if the translation gets a new dbID when it is stored.

  Initialize the merged registry with whatever is in the registry at
  that point.  Suggested by Andy Yates (thanks!).

The TranslationAdaptor now has a fetch_all_by_Transcript() method with
which it's possible to get all translations for a transcript.

The fetch_by_Transcript() method now uses the new
canonical_translation_id field in the transcript table to get the
canonical transcript.

For species without alternative initiation (all Core speaces today), the
semantics and the results are the same as before.

Using the API from now on, without patch_56_57_e.sql applied, will
result in SQL errors.

The TranscriptAdaptor now knows how to store the canonical translation ID.

results anyway).

warnings when the port is an empty string.

comparing them betwixt source and target database.