schemas/vegbien.sql: taxonconcept: Renamed canon_taxonconcept_id to canon_concept_id to shorten the name, which is used often

sql.py: mk_update(): in_place: Convert columns of type character varying to text so that they can be merge-joined with text columns. Note that these two types are equivalent but not aliases of one another, so the explicit type change is needed.

sql_gen.py: Added canon_type()

sql.py: mk_update(): in_place: Factored retrieval of column type out into separate statement for clarity

schemas/functions.sql: _join*(): Fixed bug where was returning '' instead of NULL when only NULL inputs were provided, because array_to_string() always returns a non-NULL string. Functions must always return NULL in place of '' to ensure that empty strings do not find their way into VegBIEN, and to prevent inconsistencies between row-based and column-based import (row-based import folds empty strings to NULL while column-based import relies on having a clean input table).

sql_io.py: cleanup_table(): Use sql.table_pkey_col() instead of sql.pkey_col() so that only an actual pkey column is removed from the list of columns to clean. This fixes a bug where the first column in the table was not cleaned up if there was no pkey. Note that this bug only affected newly re-created staging tables, because staging tables previously had a special row_num pkey column added if they did not already have a pkey. The row_num column is now added by column-based import instead.

sql.py: table_pkey_col(): Raise a DoesNotExistException if the table has no pkey

sql.py: pkey_col(): Call table_pkey_col() directly rather than via pkey_name(). pkey_name(): Call pkey_col() instead of table_pkey_col() now that pkey_col() calls table_pkey_col().

sql.py: pkey_col(): Documented that if there is no pkey, returns the first column in the table

sql.py: pkey_col(): Specify recover directly as a kw_arg because it's the only kw_arg passed to pkey_name()