batched coprocessor task may be sent to wrong store

[you06]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. Minimal reproduce step (Required)

1. Manually inject server-is-busy error in client-go: https://github.com/you06/client-go/blob/939d1305079f8d7a208f1ae914b043dc1092bcc4/internal/client/client.go#L692-L705
2. Build TiDB & Run cluster with 3 tikv tiup playground nightly --kv=3 --tiflash=0 --db.binpath=tidb-server
3. Run the following script

sysbench --config-file=config oltp_read_only --tables=1 --table-size=100000 --auto_inc=false prepare
mysql -h 127.0.0.1 -P 4000 -u root -D test -e "update sbtest1 set k = mod(id, 100)"
mysql -h 127.0.0.1 -P 4000 -u root -D test -e "delete from sbtest1 where mod(id, 9) != 0"
mysql -h 127.0.0.1 -P 4000 -u root -D test -e "split table sbtest1 between (0) and (100000) regions 99"
mysql -h 127.0.0.1 -P 4000 -u root -D test -e "analyze table sbtest1"


for (( i=1; i<=300; i++ ))
do
  mysql -h 127.0.0.1 -P 4000 -u root -D test -e "select * from sbtest1 force index(k_1) where k between 1 and 10" > /dev/null
  sleep 1
done

2. What did you expect to see? (Required)

No region errors or region errors in the beginning only.

3. What did you see instead (Required)

4. What is your TiDB version? (Required)

nightly.