Read a single-column CSV file as one column when detecting the format

[L4XB]

Closes #907

When format= is not given, rows_from_file() sniffs the CSV dialect. It now passes delimiters=",;\t|" to csv.Sniffer.sniff(). If none of those four characters is found, the file is read with csv.excel as a single column instead of raising csv.Error or splitting on a letter. Semicolon and tab files are detected as before. The new (b"id;name\n1;Cleo", Format.CSV) case in test_rows_from_file_detect_format pins that.

Before:

$ sqlite-utils memory names.csv "select * from names"
_csv.Error: Could not determine delimiter
$ sqlite-utils memory urls.csv "select * from urls"
[{"ur": "https://examp", "": "e.com/a"},
 {"ur": "https://examp", "": "e.org/b"}]

After:

$ sqlite-utils memory names.csv "select * from names"
[{"name": "Ann"},
 {"name": "Bob"}]
$ sqlite-utils memory urls.csv "select * from urls"
[{"url": "https://example.com/a"},
 {"url": "https://example.org/b"}]

Tests:

• test_rows_from_file_detect_format_single_column covers the two files above.
• test_memory_csv_single_column checks the same thing through sqlite-utils memory.

All three fail on main. The full test suite passes (1501), and black, flake8, mypy, pyright and cog are clean. ty reports only its existing redundant-cast warning in db.py.

insert --sniff is unchanged, since there the user explicitly asks for detection.