lib/sh/util.sh: set `shopt -s lastpipe`, to allow setting vars in the last command of a pipeline

bugfix: lib/sh/util.sh: ignore_err_msg(): use `try` properly with prep_try and `"try"`

lib/sh/util.sh: added prep_try and use it in try alias. try: documented how to run it with a wrapper command.

lib/sh/db.sh: pg_dump(): ignore "No matching tables were found" error, using new ignore_err_msg()

lib/sh/util.sh: added ignore_err_msg()

lib/sh/db.sh: pg_schema_exists(): documented that `try` is used to suppress the error exit status

lib/sh/util.sh: stderr_matches(): usage: surrounded command in if statement to indicate what context it would usually be used in

lib/runscripts/file.pg.sql.run, schema.pg.sql.run: support custom options to pg_dump in $@

lib/sql_io.py: automatic handling of input/output column type mismatches: also do this for identifying columns, which first cause an error in a join in sql.distinct_table() rather than in the main insert (and thus were not handled by the existing error handling). previously, the user would have had to manually cast the input column in postprocess.sql. this involves getting handle_MissingCastException() to update join_cols as well as mapping.

lib/sql.py: distinct_table(): don't cache the table creation, because this prevents different distinct_tables from being created for the same input table