sysfs.Mdraids() fails if any MD raid array has a delayed check or resync

If you have multiple software RAID arrays that share physical devices and you trigger a check or a resync for more than one of them, the Linux kernel MD layer notices this situation and normally delays the action on all but one device. For example:

Personalities : [raid1] [linear] [multipath] [raid0] [raid6] [raid5] [raid4] [raid10] 
md0 : active raid1 sdb2[0] sda2[1]
      972946432 blocks super 1.2 [2/2] [UU]
      [=========>...........]  check = 45.5% (442910144/972946432) finish=55.6min speed=158752K/sec
      bitmap: 2/8 pages [8KB], 65536KB chunk

md1 : active raid1 sdb3[0] sda3[1]
      3677184 blocks super 1.2 [2/2] [UU]
        resync=DELAYED
      
unused devices: <none>

When this is happening, /sys/block/md*/md/sync_completed will contain 'delayed' for any delayed arrays. This causes sysfs.Mdraids() to incorrectly fail, as the code assumes that sync_completed will contain either 'none' or '%d / %d', and attempting to parse 'completed' as the latter returns an 'expected integer' error.

Start at the sysfs.Mdraids() entry point and inspect how sync_completed values are parsed. Reproduce the delayed-array case described in the issue, then verify that delayed arrays no longer cause parsing to fail while normal none and completed/total values retain their existing behavior.