Use memrchr to find newlines in reverse JSONL scans (#49444)

Replace the byte-by-byte reverse newline search with `memchr::memrchr` and add `memchr` as a dependency of `codex-rollout`.

GitOrigin-RevId: b07e42e15efb0d63d7b00ee9a2ee29c446f4cf59

codex-rs/Cargo.lock

@@ -4432,6 +4432,7 @@ dependencies = [
  "codex-utils-absolute-path",
  "codex-utils-path",
  "codex-utils-process",
+ "memchr",
  "pretty_assertions",
  "regex",
  "serde",

codex-rs/rollout/Cargo.toml

@@ -25,6 +25,7 @@ codex-protocol = { workspace = true }
 codex-state = { workspace = true }
 codex-utils-absolute-path = { workspace = true }
 codex-utils-path = { workspace = true }
+memchr = { workspace = true }
 regex = { workspace = true }
 serde = { workspace = true, features = ["derive"] }
 serde_json = { workspace = true }

codex-rs/rollout/src/reverse_jsonl_scanner.rs

@@ -95,7 +95,7 @@ where
             }
 
             let chunk = &self.chunk[..self.chunk_position];
-            if let Some(newline_position) = chunk.iter().rposition(|byte| *byte == b'\n') {
+            if let Some(newline_position) = memchr::memrchr(b'\n', chunk) {
                 let fragment = &chunk[newline_position + 1..];
                 if !self.discarding_oversized_record {
                     if self.max_record_bytes.is_some_and(|max_record_bytes| {