block/bfq: use separate insertion lists

block/bfq: use separate insertion lists

Based on the similar patch for mq-deadline, this uses separate
insertion lists so we can defer touching dd->lock until dispatch
time.

This improves the following fio workload:

fio --bs=512 --group_reporting=1 --gtod_reduce=1 --invalidate=1 \
        --ioengine=io_uring --norandommap --runtime=60 --rw=randread \
        --thread --time_based=1 --buffered=0 --fixedbufs=1 --numjobs=32 \
        --iodepth=4 --iodepth_batch_submit=4 --iodepth_batch_complete=4 \
        --name=/dev/nvme0n1 --filename=/dev/nvme0n1

from:

/dev/nvme0n1: (groupid=0, jobs=32): err= 0: pid=1113: Fri Jan 19 20:59:26 2024
  read: IOPS=567k, BW=277MiB/s (290MB/s)(1820MiB/6575msec)
   bw (  KiB/s): min=274824, max=291156, per=100.00%, avg=283930.08, stdev=143.01, samples=416
   iops        : min=549648, max=582312, avg=567860.31, stdev=286.01, samples=416
  cpu          : usr=0.18%, sys=86.04%, ctx=866079, majf=0, minf=0
  IO depths    : 1=0.0%, 2=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0%
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     issued rwts: total=3728344,0,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=4

with 96% lock contention and 86% sys time, to:

/dev/nvme0n1: (groupid=0, jobs=32): err= 0: pid=8922: Sat Jan 20 11:16:20 2024
  read: IOPS=1550k, BW=757MiB/s (794MB/s)(19.6GiB/26471msec)
   bw (  KiB/s): min=754668, max=848896, per=100.00%, avg=775459.33, stdev=624.43, samples=1664
   iops        : min=1509336, max=1697793, avg=1550918.83, stdev=1248.87, samples=1664
  cpu          : usr=1.34%, sys=14.49%, ctx=9950560, majf=0, minf=0
  IO depths    : 1=0.0%, 2=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0%
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     issued rwts: total=41042924,0,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=4

with ~30% lock contention and 14.5% sys time, by applying the lessons
learnt with scaling mq-deadline.

Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: Jens Axboe <axboe@kernel.dk>