Catching missing Datum conversions

Hi,

When reviewing a recent patch, I missed a place where Datum was being
converted to another type implicitly (ie without going though a
DatumGetXXX() macro).  Thanks to Jeff for fixing that (commit
b538c90b), but I was curious to see if I could convince my compiler to
tell me about that sort of thing.  Here's an experimental hack that
makes Datum a struct (unfortunately defined in two places, but like I
said it's a hack), and then fixes all the resulting compile errors.
The main categories of change are:

1. Many cases of replacing "(Datum) 0" with a new macro "NullDatum"
and adjusting code that compares with 0/NULL, so you can pretty much
ignore that noise.  Likewise code that compares datums directly
without apparent knowledge of the expected type.

2.  VARDATA etc macros taking a Datum instead of a varlena *.  I think
the interface is suppose to be able to take both, so I think you can
pretty much ignore that noise too, I just couldn't immediately think
of a trick that would make that polymorphism work so I added
DatumGetPointer(x) wherever a Datum x was given directly to those
macros.

3.  Many cases of object IDs being converted implicitly, for example
in syscache calls.  A few cases of implicit use as booleans.

4.  Various confusions about the types involved in PG_RETURN_XXX and
PG_GETARGS_XXX macros, and failure to convert values from
Datum-returning functions, or unnecessary conversion of results (eg
makeArrayResult).

I should probably split this into "actionable" (categories 3 and 4)
and "noise and scaffolding" patches.

-- 
Thomas Munro
https://enterprisedb.com