[keichi]

Problem

generateString had two bugs when both length and zeroTerminated are set:

1. If the NUL terminator is exactly the last byte within length, it was included in the result, because the end computation could not distinguish "hit NUL at the last byte" from "hit the length limit".
2. length was inlined raw instead of going through ctx.generateOption, so function lengths caused a SyntaxError and field-name lengths became bare (undefined) identifiers.

Repro

new Parser().string("s", { length: 4, zeroTerminated: true }).parse(Buffer.from("abc\0")).s
// "abc\u0000" (expected "abc")

new Parser().uint8("n").string("s", { length: "n", zeroTerminated: true }) // ReferenceError at parse
new Parser().string("s", { length: () => 4, zeroTerminated: true })       // SyntaxError

Fix

Evaluate the length once via ctx.generateOption into a tmp variable, scan until the limit or a NUL (without consuming it), decode [start, offset), then skip the NUL if one was found. Offset semantics are unchanged (advances just past the NUL, or by length if none). Added regression tests.