new queries (v3?)
-----------------

pseudo-design-note:

 # id based queries removed completly
 # basic hashing scheme
   * name of a query is hashed,
   * part of a hash is used to select one of QUERIES_BUCKETS buckets,
   * each bucket is a list holding queries
   > we could employ some more sophisticated hashing scheme,
     but since number of queries (that is unique names)
     is relatievly low, there is no need for that

 > because this is experimental code, I've left a call
   to xstrcmp(), but we should probably disallow registering
   queries that have colliding hashes,
   (read: we should select such a hash function, that will be
    both fast and won't cause collisions, at least on queries
    used by plugins distributed with ekg2)
   So later we should be able to get rid of xstrcmp() call.

kind-of-a-rationale:

 it seems that changes related to id-based queries made
 more bad than good:
 + known-to-core queries are fast (constant lookup time)
 - id-based queries makes the code more complicated
 - problem for script languages


api:

query_register - register given query

query_connect - attach a handler to the given query,
        you probably shouldn't connect if the query hasn't
        yet been registered

query_emit - emit a query


