Description

The current CSV parser splits on commas and trims line endings, but does not handle the full RFC 4180 quoted-field syntax. Fields that contain commas, double quotes, or embedded newlines must be enclosed in double quotes — this is extremely common in real-world data.

Acceptance Criteria

• Fields enclosed in "..." are parsed correctly as a single value
• Embedded commas inside quoted fields are not treated as delimiters
• Escaped double quotes ("") inside quoted fields are decoded to "
• Fields with embedded newlines (\n) are read as a single multi-line value
• Existing behaviour for unquoted fields is unchanged
• At least 5 unit tests covering edge cases
• Parser is single-pass and streaming — no buffering the full input (stdin cannot seek)

Notes

• Reference: RFC 4180
• Should handle both \n and \r\n line endings within quoted fields
• Must use a single-pass streaming parser — this is required (not optional) because sql-pipe reads from stdin, which is not seekable. Buffering the entire file is not acceptable for large inputs.
• Implement as a state machine over individual bytes/codepoints

Refinement note (Sprint 1): Clarified that streaming/single-pass is a hard requirement, not a suggestion. Moved it to AC.