LU-17865 osc: fiemap deadlock fix

LU-17865 osc: fiemap deadlock fix

A fiemap call may deadlock due to wrongly requesting an ldlm lock at
server while the same lock is cached and pinned at the client. Two PR
lock requests are compatible so the deadlock also needs a concurrent
write lock.

ll_fiemap_info_key is shared between osc_object_fiemap()
calls, once OBD_FL_SRVLOCK flag is set, it is reused for
all subsequent RPCs regardless of the local lock caching status.

HPE-bug-id: LUS-12353
Signed-off-by: Alexander Zarochentsev <alexander.zarochentsev@hpe.com>
Change-Id: I6e76bc5e4549ed887b8f6177432acf90f9ec614d
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/55163
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andrew Perepechko <andrew.perepechko@hpe.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>