inputs/.TNRS/tnrs/map.csv: Added _nullIf filter to remove "Unknown" values for Accepted_name_family

README.TXT: Generate the local TNRS cache from the test_taxonomic_names rather than syncing it with the vegbiendev TNRS cache, so that the automated test's inserted row count stays the same regardless of the contents of the full-DB TNRS cache

README.TXT: Backups: Added TNRS cache section

inputs/.TNRS/tnrs/test.xml.ref: Accepted inserted row count using TNRS cache created from test_taxonomic_names. Using a standard set of names for the test ensures that the inserted row count will not change when the full-DB TNRS cache changes.

inputs/.TNRS/schema.sql: tnrs_accepted_names: Prepend the Accepted_name_family to the taxonomic name that will be submitted back to TNRS for parsing, because TNRS input names now always include the family when it's provided

inputs/.TNRS/schema.sql: tnrs_accepted_names: Use simpler array_to_string() instead of || and COALESCE to put together the taxonomic name that will be submitted back to TNRS for parsing. Note that this requires defining an IMMUTABLE wrapper function for array_to_string(), because pg_catalog.array_to_string() is declared STABLE but indexes require functions to be IMMUTABLE (http://www.mail-archive.com/pgsql-hackers@postgresql.org/msg156323.html).

inputs/.TNRS/schema.sql: Don't hardcode the schema name

input.Makefile: Staging tables installation: sql/install: Provide the datasource's schema to the script in :schema, so it can refer to its own elements explicitly when it's not possible to rely on the search_path. This is the case for functions that have the same signature as (and are intended to replace) a pg_catalog function, because the pg_catalog function will be used in preference to the datasource function regardless of the search_path.

input.Makefile: Staging tables installation: $(cleanup): If a cleanup.sql is provided, only run it and don't do default cleanup, to allow tables to override rather than just add to default cleanup operations. This prevents the automatic replacement of certain strings (sql_io.null_strs) with NULL on TNRS, and keeps the TNRS cache mostly as it was output by the TNRS service. Note that empty strings are still replaced with NULL by COPY FROM in sql_io.append_csv(). This is necessary for TNRS import to work properly, because although '' generally means NULL, it is not treated that way by PostgreSQL.

input.Makefile: Staging tables installation: Moved custom cleanup.sql cleanup operations to main $(cleanup) function, so custom cleanup operations would run whenever any target (such as %/install) invokes $(cleanup), not just manually through %/cleanup