avoid queuing transfers for remotes after syncing to them is paused

This avoids the expensive transfer scan relying on its list of remotes
to scan being accurate throughout, which it will not be when the user
pauses syncing to a remote.

I feel it's ok to queue transfers to *any* known remote, not just the ones
being scanned.

Note that there are still small races where after syncing to a remote is
paused, a transfer can be queued for it. Not just in the expensive transfer
scan, but in the cheap failed transfer scan, and elsewhere.

Assistant/Threads/TransferScanner.hs

@@ -87,7 +87,7 @@ failedTransferScan st dstatus transferqueue r = do
 			transferqueue dstatus (associatedFile info) t r
 
 {- This is a expensive scan through the full git work tree, finding
- - files to download from or upload to any of the remotes.
+ - files to download from or upload to any known remote.
  - 
  - The scan is blocked when the transfer queue gets too large. -}
 expensiveScan :: ThreadState -> DaemonStatusHandle -> TransferQueue -> [Remote] -> IO ()
@@ -114,7 +114,11 @@ expensiveScan st dstatus transferqueue rs = unless onlyweb $ do
 			queueTransferWhenSmall transferqueue dstatus (Just f) t r
 		findtransfers f (key, _) = do
 			locs <- loggedLocations key
-			let use a = return $ catMaybes $ map (a key locs) rs
+			{- Queue transfers from any known remote. The known
+			 - remotes may have changed since this scan began. -}
+			let use a = do
+				knownrs <- liftIO $ knownRemotes <$> getDaemonStatus dstatus
+				return $ catMaybes $ map (a key locs) knownrs
 			ifM (inAnnex key)
 				( filterM (wantSend (Just f) . Remote.uuid . fst)
 					=<< use (check Upload False)