fix: Use in-place append for buffer concatenation to fix O(n^2) performance

[keelerm84]

The previous change (0e2e80d) replaced buffer << chunk with
buffer += chunk to handle frozen strings. However, += creates a new
string on every iteration, resulting in O(n^2) behavior for large
payloads. The buffer is already explicitly unfrozen (+"".b), so <<
is safe to use. This restores the in-place append while keeping the
.dup on the incoming chunk to avoid mutating frozen caller strings.

---

Note

Low Risk
Low risk: small, localized change to streaming buffer concatenation; main concern is any subtle string mutability/encoding edge cases during line parsing.

Overview
Restores in-place buffer growth in BufferedLineReader.lines_from by switching from buffer += chunk back to buffer << chunk, avoiding per-chunk string allocations and O(n^2) behavior on large streams.

Also ensures the buffer reset path reinitializes to an unfrozen binary string (+"".b) rather than a potentially frozen empty literal.

^{Written by Cursor Bugbot for commit 37d5f65. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}