Reverse records separated by a single byte.
Uses forward SIMD scan (memchr_iter) to collect all separator positions,
then fills output buffer in reverse order with parallel copy for large data.
Single write_all at the end for minimum syscall overhead.
Reverse records using a regex separator.
Uses regex::bytes for direct byte-level matching (no UTF-8 conversion needed).
NOTE: GNU tac uses POSIX Basic Regular Expressions (BRE), so we convert to ERE first.
Uses backward scanning to match GNU tac’s re_search behavior.