use LIKE.

+  Arg [2]       : (optional) String
+                  The subset within the ontolgy to which the query
+                  should be restricted.  The subset may be specified
+                  as a SQL pattern, e.g., "goslim%".
+
+  Arg [3]       : (optional) Boolean
+                  If true (non-zero), only return the closest
+                  terms.  If this argument is true, and the previous
+                  argument is left undefined, this method will
+                  return the parents of the given term.
+
+  Description   : Given a child ontology term, returns a list of
+                  all its ancestor terms, up to and including any
+                  root term.  Relations of the type 'is_a' and
+                  'part_of' are followed.  Optionally, only terms in
+                  a given subset of the ontology may be returned,
+                  and additionally one may ask to only get the
+                  closest term(s) to the given child term.

Hopefully speeds up the loading of the term table a bit, especially over
a slow link (SSH from home).

number of separate variables.

'term' table.  The SET is hard-coded and needs updating if new subsets
are added :-/ No index yet as I don't know how we're going to use it.

with non-default names.  Unbreaks health-checking.

from Will.  Unbreaks health-checks for them.

inherited and used by the Variation, Funcgen, or Compara APIs, because
these other schemas didn't use to have a species_id column in their meta
tables.  Now they do, so we're removing the exceptions.

to merge copies of it; added a _SPECIES and a _TYPE section to the hash.

Added the subroutine load_registry_from_multiple_dbs():

   Arg [1]   : Array of hashes, each hash being a set of arguments to
               load_registry_from_db() (see above).

   Example   :

     $registry->load_registry_from_multiple_dbs( {
         '-host'    => 'ensembldb.ensembl.org',
         '-user'    => 'anonymous',
         '-verbose' => '1'
       },
       {
         '-host'     => 'server.example.com',
         '-user'     => 'anonymouse',
         '-password' => 'cheese',
         '-verbose'  => '1'
       } );

   Description:  Will call load_registry_from_db() (see above)
                 multiple times and merge the resulting registries
                 into one, effectively allowing a user to connect to
                 databases on multiple database servers from within
                 one program.

                 If a database is found on more than one server, the
                 first found instance of that database will be used.

cache is used.

  Add the parameters bound with bind_param_generic_fetch() to the cache
  key.

  Also some rearrangement of logic.

Now choose which dumping method to use depending on the number of genes and the number of toplevel sequences