mtd: ubi: Fix worker handling

mtd: ubi: Fix worker handling

Fixes a bug found on thuban boards, which were for 2 years in
a long-term test with varying temperatures. They showed
problems in u-boot when attaching the ubi partition:

U-Boot# run flash_self_test
Booting from nand
set A...
UBI: attaching mtd1 to ubi0
UBI: scanning is finished
data abort
pc : [<87f97c3c>]          lr : [<87f97c28>]
reloc pc : [<8012cc3c>]    lr : [<8012cc28>]
sp : 85f686e8  ip : 00000020     fp : 000001f7
r10: 8605ce40  r9 : 85f68ef8     r8 : 0001f000
r7 : 00000001  r6 : 00000006     r5 : 0001f000  r4 : 85f6ecc0
r3 : 00000000  r2 : 44e35000     r1 : 87fcbcd4  r0 : 87fc755b
Flags: nZCv  IRQs off  FIQs on  Mode SVC_32
Resetting CPU ...

Reason is, that accidentially the U-Boot implementation
from __schedule_ubi_work() did not check the flag
ubi->thread_enabled and started with wearleveling work,
but ubi did not have setup all structures at this point
and crashes.

Solve this problem by splitting work scheduling and processing.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Heiko Schocher <hs@denx.de>