strings.py: concat(): Convert args to raw (non-Unicode) strings first, so that multi-byte Unicode sequences are considered by # of bytes instead of # of chars. This is necessary because PostgreSQL truncates identifiers by # of bytes instead of # of chars, so that identifiers will actually be less than 63 chars long when some chars were multi-byte.

strings.py: ustr(): Call str() method manually like urepr() to avoid Unicode errors when the returning string is non-ASCII

strings.py: Added urepr() and use it in repr_no_u(), to better support repr() return values with non-ASCII characters. Avoiding repr() also provides a more complete stack trace in the case of such errors.

schemas/vegbien.sql: plantobservation: plantobservation_aggregateoccurrence_count_1() trigger: Don't raise an error if existing count was >1, because there are in fact datasets (notably SALVIAS) where input records for individual stems may themselves contain aggregate data (such as plant and stem counts). For this data, we have an anomalous condition where an aggregateoccurrence has count >1 but contains one plantobservation, due to the plant/stem count being included in the first stem's record. (See <https://projects.nceas.ucsb.edu/nceas/projects/bien/wiki/SALVIAS_issues#Data-interpretation-issues> for more info on this problem.) Note that our desired 1:1 relationship between aggregateoccurrence and plantobservation is still guaranteed by a constraint, but the anomalous data may still cause irregularities later on in the analysis.

sql_io.py: put_table(): Ignoring all rows on unrecoverable errors: Also support the case where has_joins == True, by setting it to False so that the no-joins case is effectively used

inputs/import.stats.xls: Moved Simultaneously above Independently because that is how we are now running the imports

Regenerated vegbien.ERD exports

schemas/vegbien.sql: 1_to_1 and *_unique_within unique indexes with a `WHERE sourceaccessioncode IS NULL` filter: Added IS NULL filters for other unique keys, so that these fallback indexes would only be used if there was no (or no other) way to uniquely identify their tables. For *_1_to_1 unique indexes, this is the case for specimens data.

schemas/vegbien.sql: 1_to_1 and *_unique_within unique indexes with a `WHERE sourceaccessioncode IS NULL` filter: Added IS NULL filters for other unique keys, so that these fallback indexes would only be used if there was no (or no other) way to uniquely identify their tables. For *_1_to_1 unique indexes, this is the case for specimens data.

schemas/vegbien.sql: stemobservation: Replaced stemobservation_unique_code unique constraint with stemobservation_unique_within_plantobservation unique index that uses COALESCE and WHERE ... IS NOT NULL appropriately, to work with sql_gen's use of COALESCE indexes and (for the renaming) to better reflect what it does