[pauloflp]

With rtrim: true or trim: true, whitespace after a closing quote is recognized but only its first byte is skipped. The remaining bytes then trigger CSV_NON_TRIMABLE_CHAR_AFTER_CLOSING_QUOTE.

For example, parse(Buffer.from('"a"\u3000,"b"'), { trim: true }) currently throws instead of returning [['a', 'b']]. UTF-16LE also fails with ordinary spaces because they occupy two bytes.

Advance past the complete matched trim character when discarding trailing whitespace, as the leading-whitespace path already does. Add regressions for UTF-8 and UTF-16LE, both options, empty quoted fields, field/record boundaries and EOF, preservation of whitespace inside quotes, one-byte stream writes, and rejection of a non-whitespace character after the padding.

Validation on Node.js v24.22.1:

• The new tests produce 8 failures before the source fix.
• npm run test --workspace=csv-parse: 641 passing, 2 pending; TypeScript check passes.
• ESLint on both changed files passes.
• npm run build --workspace=csv-parse passes.

Related to the whitespace support introduced in #483. This addresses consuming already-recognized trim characters after quotes; it does not change the BOM state refresh in #537. The fixtures specify their record delimiter to keep this separate from #539.

AI-assisted implementation and local verification.