Reduce allocations in PulsarEntryFormatter batch encoding

[Denovo1998]

Motivation

PulsarEntryFormatter currently builds an intermediate list of converted messages before serializing the final batch payload, which adds extra allocations and increases peak memory usage. In addition, the batch buffer starts with a small fixed size and is only released on the normal return path, which can lead to unnecessary buffer growth and potential ByteBuf leaks if an exception occurs. Streamlining the encode path reduces overhead and makes the transactional/control metadata handling easier to follow.

Modifications

• Pre-size the batch buffer using AppendInfo.validBytes() (bounded by a maximum) to reduce reallocations and copies.
• Serialize records into the batch payload in a single pass instead of materializing an intermediate message list.
• Cache isTransactional / isControlBatch per record batch to reduce repeated checks and improve readability.
• Guard batch size tracking/logging work behind log.isTraceEnabled() to avoid extra overhead in normal operation.
• Always release the temporary batch buffer via try/finally to prevent leaks on error paths.