plpgsql: optimize "SELECT simple-expression INTO var".

plpgsql: optimize "SELECT simple-expression INTO var".

Previously, we always fed SELECT ... INTO to the SPI machinery.
While that works for all cases, it's a great deal slower than
the otherwise-equivalent "var := expression" if the expression
is "simple" and the INTO target is a single variable.  Users
coming from MSSQL or T_SQL are likely to be surprised by this;
they are used to writing SELECT ... INTO since there is no
"var := expression" syntax in those dialects.  Hence, check for
a simple expression and use the faster code path if possible.

(Here, "simple" means whatever exec_is_simple_query accepts,
which basically means "SELECT scalar-expression" without any
input tables, aggregates, qual clauses, etc.)

This optimization is not entirely transparent.  Notably, one of
the reasons it's faster is that the hooks that pg_stat_statements
uses aren't called in this path, so that the evaluated expression
no longer appears in pg_stat_statements output as it did before.
There may be some other minor behavioral changes too, although
I tried hard to make error reporting look the same.  Hopefully,
none of them are significant enough to not be acceptable as
routine changes in a PG major version.

Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Pavel Stehule <pavel.stehule@gmail.com>
Discussion: https://postgr.es/m/CAFj8pRDieSQOPDHD_svvR75875uRejS9cN87FoAC3iXMXS1saQ@mail.gmail.com