[improvement](tablet clone) furthur repair replicas should be check even if they are versions catchup

[yujun777]

Proposed changes

pick: #25551

If there are loading txns, PR #24845 will try to furthur repair a balance tablet even if their versions are catchup.

Then in scheduler's NEED_FURTHER_REPAIR handler, if no replicas are version fallbehind, the sched ctx will convert to REPLICA_MISSING, then add missing replicas will fail because they don't really miss replicas.

Improvement: for NEED_FURTHER_REPAIR handler, if all the replicas are version catchup, then try to check the txns before further repair txn has finished, if it finished, then mark the replica as further repair finish.

Further comments

If this is a relatively large or complex change, kick off the discussion at dev@doris.apache.org by explaining why you chose the solution you did and what alternatives you considered, etc...

[yujun777]

run buildall

[doris-robot]

(From new machine)TeamCity pipeline, clickbench performance test result:
the sum of best hot time: 48.84 seconds
stream load tsv: 558 seconds loaded 74807831229 Bytes, about 127 MB/s
stream load json: 20 seconds loaded 2358488459 Bytes, about 112 MB/s
stream load orc: 65 seconds loaded 1101869774 Bytes, about 16 MB/s
stream load parquet: 31 seconds loaded 861443392 Bytes, about 26 MB/s
insert into select: 29.4 seconds inserted 10000000 Rows, about 340K ops/s
storage size: 17162644213 Bytes

[xiaokang]

run buildall

[doris-robot]

(From new machine)TeamCity pipeline, clickbench performance test result:
the sum of best hot time: 50.59 seconds
stream load tsv: 559 seconds loaded 74807831229 Bytes, about 127 MB/s
stream load json: 20 seconds loaded 2358488459 Bytes, about 112 MB/s
stream load orc: 65 seconds loaded 1101869774 Bytes, about 16 MB/s
stream load parquet: 32 seconds loaded 861443392 Bytes, about 25 MB/s
insert into select: 29.5 seconds inserted 10000000 Rows, about 338K ops/s
storage size: 17162115028 Bytes

[xiaokang]

LGTM