This reduces the MonetDB error rate from 90% to 50%.  MonetDB doesn't
appear to support LIMIT clauses in atomic subselects.

The previous commit surprisingly yielded a CPU hogging sqlsmith with
monetdb due to the table_or_query_name production hardly finding
candidates.  It turned out that the scope model was seriously lacking:
Creating a child scope now automatically includes the table and column
references available to the parent scope.  Previously, child scopes
only contained table and column references explicitly put there by the
parent production.

This special-casing for MonetDB is no longer needed as the impedance
matching code will blacklist the production.

Throw the correct exception classes and fill them with sensible data.
Also avoid writing anything to stdout from the DUT class.  If the
previous, extremely verbose output is desired, a logger class should
be implemented that is made available with a new option, e.g.,
--log-to-file.

Some messages report back huge literals constructed by sqlsmith.
There's no use dumping these on the terminal.

Add pkg-config to brew install command to avoid this error:
checking for PostgreSQL libraries... yes
./configure: line 4064: syntax error near unexpected token `LIBPQXX,'
./configure: line 4064: `PKG_CHECK_MODULES(LIBPQXX, libpqxx >= 4.0)'

Reported by @philwinn.

This gets rid of the reset_gucs hack.

Allows starting up with a previously serialized rng state.

Errors are properly handled and reported on non-generated statements
as well now.  Also gets rid of connection handle leak.

This gets rid of the regexp-matview-optimization.

Some instances died with "basic_string::_M_construct null not valid".
Maybe it happened here.

This allows access to the sqlstate and ought to get rid of the
confused transaction state bug.

Closes #12, closes #13.

By materializing a single regexp from all lines in known_re, the CPU
load by filtering goes down by 90%.