lib/sh/db.sh: set test mode when using limited # rows

bugfix: inputs/GBIF/table.run: table.tsv.gz/upload(): don't use inplace mode because it leaves a newer mtime when aborted, causing rsync to think that the partial upload is actually newer than the source. note that rsync's --partial-dir mode is just as capable of resuming an aborted upload (it will just use a file in .rsync-tmp instead). inplace mode is primarily designed for fixed-offset files which don't change much between edits, but this is not true for exports (or the gzips of them), which will change the file offsets of most data if even one row or column is added or removed.

bugfix: inputs/GBIF/table.run: table.tsv.gz/make(): run table.tsv.gz/upload here instead of in table.tsv/make() because it should not run until table.tsv.gz is finished being made, which is not the case in table.tsv/make() because table.tsv.gz/make is run in the background

inputs/GBIF/table.run: table.tsv.gz/upload(): moved before table.tsv.gz/make() so it can be used by it

bugfix: inputs/GBIF/table.run: table.tsv.gz/upload(): need overwrite=1 because the mtime of an aborted inplace upload is newer

inputs/GBIF/table.run: table.tsv*/upload(): renamed to table.tsv.gz/upload() to upload only table.tsv.gz, not table.tsv, in order to save bandwidth

bugfix: lib/sh/sync.sh: also need to --include parent dirs for each --include path

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

*{.sh,run}: in comments, use ${array[@]} instead of @array for clarity

lib/sh/util.sh: foreach_arg(): moved `local a` to same line as for loop that uses it