Further test stabilisation

 - ndbinfo_cluster_locks
   - SCAN vs other lock types - scan locks will be converted
     to READ locks by takeover ops, and whether these are issued
     or not depends on timing, data distribution etc.
     Therefore, the type of the locks is ignored in two affected
     testcases.

 - ndbinfo_cluster_locks + ndbinfo_locks_per_fragment
   - Testcases where sine kind of 'successful wait' is to be
     observed : Slow platforms turn this into transaction
     timeout.
     Solution : Disable transaction timeout where we
     do not want it.  This involved enhancing an existing
     DUMP code in DbtcMain to allow it to 'reset' the
     timeout to the configured value.

 - ndbinfo
   - Remove existing testcase side-effect of dropping ndb$test
     as this makes test 1-shot only, not suitable for use with
     mtr --repeat

 Also : Remove missing newline at end of ndbinfo_init_frag_lock_count.inc
 which meant it could cause parse issues in the including
 .test file.