Fix bugs around compaction

[zipper-meng]

Required checklist

• Sample config files updated (config,meta/config and default config)
• If there are user-facing changes, the documentation needs to be updated prior to approving the PR( Link )
• If there are any breaking changes to public APIs, please add the api change label.
• Signed CLA (if not already signed)

Which issue does this PR close?

Related #.

Rationale for this change

Fix some bugs:

1. Compaction job failed to start;

CompactJobInner::enable_compaction is initialized as true but the job starts only when it compare_exchange(false, true) == Ok(false).

cnosdb/tskv/src/compaction/job.rs

Lines 122 to 134 in 3112894

	if self
	.enable_compaction
	.compare_exchange(
	false,
	true,
	atomic::Ordering::Acquire,
	atomic::Ordering::Relaxed,
	)
	.is_err()
	{
	info!("Compaction: enable_compaction is false, later to start vnode compaction job");
	return;
	}

2. Errors in compact::chunk_merged_block();

There may be an error when chunking big blocks, this makes data lost.

3. Merge two tmp-fields in CompactIterator for better performance;

Merge tmp_tsm_blk_meta_iters: Vec<BlockMetaIterator> and tmp_tsm_blk_tsm_reader_idx: Vec<usize> into tmp_tsm_blk_meta_iters: Vec<(usize, BlockMetaIterator)>.

4. Tombstone files not removed when removing TSM file

Tombstone files may be removed when dropping a ColumnFile.

Are there any user-facing changes?