Fix CSV columnizer crash when "first line contains field names" is unchecked (#707)

[Hirogen]

Fixes #707

Problem

• Unchecking "first line contains field names" in the CSV columnizer config showed an error dialog, and after that nothing was split into columns. CsvHelper's ReadHeader() throws ReaderException: Configuration.HasHeaderRecord is false., and the columnizer called it for every line, including when selecting the columnizer (Selected()).
• Column names for files without a header were built by gluing strings together, giving Column 01, Column 11, …
• Found while testing the fix: the columnizer instance being selected or configured is often a clone that loaded csvcolumnizer.json and never ran PreProcessLine() on the file. It kept the saved delimiter (e.g. , from an earlier comma file), not the file's own. A quoted ; file parsed with , is bad CSV data, so Selected() threw BadDataException. This happened with or without a header.

Changes (CsvColumnizer)

• Read fields directly: a new ReadFields() helper reads each line's fields from csv.Parser.Record instead of the header API. Selected() and SplitCsvLine() both use it. Behavior for files with a header is unchanged (HeaderRecord was just Parser.Record).
• Column names: files without a header get Column 1 … Column N.
• Bad first line: if the first line can't be parsed with the current settings, Selected() falls back to a single Text column, as SplitCsvLine() already does. There's no exception dialog anymore.
• Delimiter detection:
  • Selected() detects the delimiter from the file's first line, as PreProcessLine() does when a file loads.
  • Configure() detects before opening the dialog, so the dialog shows the file's delimiter.
  • After OK, the columns are rebuilt without detecting again, so a delimiter typed by the user is kept.

Tests

New tests in CSVColumnizerTest:

• No header: Selected() gives 7 columns named Column 1–Column 7, SplitLine() returns the 7 field values, and the first line stays visible.
• Saved , delimiter on a ; file: fields are split correctly without a header, and column names come from the header with one.
• A first line that can't be parsed falls back to Text in both header modes.

Known limitations (not changed)

• A delimiter typed in the dialog is replaced by the detected one when the file reloads (existing PreProcessLine() behavior).
• Configure() still saves the detected delimiter to csvcolumnizer.json.
• After switching header mode, the header row only reappears as data once the file reloads.