CancelAllBackgroundWork: Flush of unpersisted data waits for stall conditions to clear and delays db close

Hey RocksDB team,

I have a small question regarding the behavior when closing the db.
Currently, when `CancelAllBackgroundWork` is called, a flush is initiated if there is any unpersisted data - here https://github.com/facebook/rocksdb/blob/acf77e1bfee9ebb0867ac277927ed2e37276c493/db/db_impl/db_impl.cc#L496
The above flush request will wait until certain stall conditions are cleared - in detail here: https://github.com/facebook/rocksdb/blob/acf77e1bfee9ebb0867ac277927ed2e37276c493/db/db_impl/db_impl_compaction_flush.cc#L2617
Since the default value for `allow_write_stall` in `FlushOptions` is false. This means the DB Close() will have to wait for those stall conditions to clear.

I find this behaviour unexpected since the user is sometimes unaware of this connection and the call to close the db might take much longer than expected.
Is there a reason why we should not pass `allow_write_stall = true` ?

`CancelAllBackgroundWork(true)` can be passed if waiting for bg work is needed and the problem as i see it currently is that even when calling `CancelAllBackgroundWork(false)`, the call will be delayed in cases of unpersisted data and a stall condition.

Thanks