TEST RESULTS SHOULD NOT DEPEND ON NUMBER OF TC / LDM THREADS IN CONFIG

Several MTR tests are modified such that the expected *.result 
file is independent of how many LDM / TC threads are configured, 
and on whether the ndbd or ndbmtd is used.

Most changes are related to default 'partiotions' being a function
of number of LDM's. Thus explicit 'partiotions <n>' spec has been 
added to several create table statements. 

A few places the test itself tests that the expected number of
default partitions are created. In these places we have 
included 'ThreadConfig=ldm={count=1}' in the *.cnf file.

Also had to increase setting of some config variable in order to 
account for possible more partitions being created with bigger
configs.

Also fixed a testcase for bug#58280 (ndb_index.test), which
required that 'index_merge' access method was used to test
aganst a previous defect. (Added more non-matching rows to
test table to make 'index_merge' more favourable)

In the join_pushdown test, I have now removed verification of
some 'info_counters'. These has in the past caused several
issues due to failing test, and maintaining these are IMO not
worth the effort anymore.