LU-15282 lod: less spinlock on the alloc rr

LU-15282 lod: less spinlock on the alloc rr

Don't need to hold spinlock for so much time, anyway it's released
in middle of loop, so RR cannot be perfect in multithreaded case.

Fix small bug in RR precession for stripecount=4+OSTCOUNT=6.

Lustre-change: https://review.whamcloud.com/45694
Lustre-commit: fa34a15f619c49535e214eec0d076872ac9a6b9a

Fixes: 665e36b780f ("OST pools on HEAD")
HPe-bug-id: LUS-10627
Signed-off-by: Alexey Lyashkov <alexey.lyashkov@hpe.com>
Change-Id: I66eded451c8256de0e5a9a0eb862af8b306da9e1
Reviewed-by: Alexander Boyko <alexander.boyko@hpe.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-on: https://review.whamcloud.com/c/ex/lustre-release/+/49757
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>