Fix a bug in AbortIO

[MJY-HUST]

Summary: When using asynchronous IO based on io_uring, if the io_handles of the cancelled request are not recorded when canceling the request within AbortIO, it may cause other asynchronous IO requests to be processed incorrectly, resulting in the program getting stuck.
Reproduction:
fill data:

./db_bench -db=/tmp/my_db -benchmarks="fillseqdeterministic" -key_size=32 -value_size=512 -num=5000000 -num_levels=4 -multiread_batched=true -use_direct_reads=false -adaptive_readahead=true -threads=1 -async_io=false -multiread_stride=40000 -disable_auto_compactions=true -compaction_style=1 -bloom_bits=10

read data: seek with next

./db_bench -use_existing_db=true -db=/tmp/my_db -benchmarks="seekrandom" -key_size=32 -value_size=512 -num=5000000 -batch_size=8 -multiread_batched=true -use_direct_reads=true -reads=1000 -ops_between_duration_checks=1 -readonly=true -threads=1 -cache_size=0 -async_io=true -multiread_stride=10000 -statistics -adaptive_readahead=true -bloom_bits=10 -seek_nexts=10000

Test: Successfully executed the above asynchronous query benchmark