dynchk and mkheader generators need to report more when problems found

Several times when new data have been added to dynchk, dynchk library
generation fails (some cases were mentioned in bug 3680, although those
specific cases are fixed).

dynchk's gen_lib.pl is not in error here - in each case that's been pursued we
find the data are somewhat incomplete, or in error. However, gen_lib does not
give much help in tracking these down, we get Perl errors on use of
uninitialized variables, but no hint what was being done. It would be better
to detect the error condition and print a diagnostic than to let the Perl
interpreter complain - because at that point we're not in control of the output
and can't show context.

There's also one function where data are clearly bad because we get a
program-generated output:

empties on Type 0; Param number -1 in Interface 0

at this point we also have no real context, this is in get_type_string(), but a
DB query is done based on the passed parameter $type_id, the value of which is
printed as Type in the message above - it clearly wasn't useful to proceed even
this far, the DB query won't succeed as there is not Type.Tid=0. As the type
we're querying is invalid, we can't print any information here about what
actually was wanted, that would have to be checked for at the point this
routine is called.

A copy of devel/misc-test/dynchk/libs/gen_lib.pl is being committed together
with this bug annotating the currently known error locations, as noted near the
top of the file these are marked with BUG.