For #439.
`file_progress()` sets the progress bar length to the file size in
bytes (`os.path.getsize(file.name)`), but `UpdateWrapper.__iter__`
called `update(len(line))`, which is the decoded character count.
With UTF-16-LE input every character is 2 bytes, so the bar capped
at 50%; UTF-32 capped at 25%; etc.
Simon noted in the issue that the obvious fix (calling `.tell()` on
the wrapped text stream) doesn't work because text mode disables it
during iteration. The underlying binary buffer doesn't have that
restriction though, so this tracks progress against
`TextIOWrapper.buffer.tell()` when the wrapped object exposes one.
For raw binary streams (no `.buffer` attribute) we keep the old
behaviour, which was already byte-accurate.
Added six regression tests in tests/test_utils.py covering UTF-8,
UTF-16-LE, BOM-prefixed UTF-16, the sniff-style `BufferedReader` chain,
a raw binary fallback, and the `.read()` path used by the JSON loader.
Each asserts that the sum of update() calls equals the on-disk file
size, which is what `click.progressbar` needs to reach 100%.
Queries returning duplicate column names - e.g. joins between tables
sharing column names - silently lost values because rows were built
with dict(zip(keys, row)), where the last duplicate wins.
Later occurrences are now renamed with a numeric suffix: id, id
becomes id, id_2 - skipping any suffix that would collide with a
real column in the same query.
The new utils.dedupe_keys() helper transforms the key list once per
query, so the per-row dict construction is unchanged and there is no
measurable performance impact.
Applied in Database.query() (including the PRAGMA and RETURNING
paths), Table.rows_where(), Table.search() and the CLI's JSON output.
CSV, TSV and table output keep the original duplicate headers.
Closes#624
