Support replaying audit logs

[djshow832]

Feature Request

Traffic files are generated by TiProxy but many clusters don't have TiProxy deployed, so we can support replaying from audit logs.

• Support reading from TiDB audit logs #859
• Deduplicate audit logs and ignore uninitialized statements #862
• Support reading audit log files in order #864
• Support replaying prepared statements #867
• Update prepared statement ID in the replay command #870
• Support replaying with a specified start time #874
• Support reading multiple traffic paths concurrently #875
• Replay in the client mode #876
• Support ignoring errors in the replayer #889
• Decode audit logs in advance #888
• List S3 file with prefix #892
• Add logs for replay progress #904
• Support --ps-close=never mode #922
• Reading S3 and decoding commands in parallel #940
• Add more fields in the report tables #929
• Support graceful shutdown of the replayer #927
• Support dry-run mode in traffic replay #938
• Record and load the timestamp from disk for replay #950
• Filter the commands by command end time #959
• Init DB when replay instead of parsing #964
• Automatically detect new TiDB audit log path #968
• Output replayed SQL into log files #974
• Support start replayer in service mode #970
• Add a new argument wait-until to start the replay job later #1001
• Support no auto quit mode for replayer #1004
• Connect to itself in the dry-run mode #969
• Output the heap profile when the memory usage is high #979
• Limit the decoded commands within a memory usage
• Filter retried audit logs by the RETRY field #1008
• Support comparing execution time and affected rows from audit logs and group by SQL digest #1014