flexnbd serve: Make the killswitch per-client-thread

This is a bit tricky, but calling shutdown() on a socket in a signal
handler is safe, and (at least in linux) appears to cause any read()
or write() calls blocked on that socket to return, even with SA_RESTART.

I'm not confident enough about the rest of flexnbd's syscall error
handling to turn SA_RESTART off for this signal...

README.txt

@@ -25,7 +25,7 @@ COMMANDS
 serve
 ~~~~~
   $ flexnbd serve --addr <ADDR> --port <PORT> --file <FILE>
-    [--sock <SOCK>] [--default-deny] [global option]* [acl entry]*
+    [--sock <SOCK>] [--default-deny] [-k] [global option]* [acl entry]*
 
 Serve a file. If any ACL entries are given (which should be IP
 addresses), only those clients listed will be permitted to connect.
@@ -55,6 +55,12 @@ Options
     empty ACL will let no clients connect.  If it is not given, an
     empty ACL will let any client connect.
 
+*--killswitch, -k*:
+    If set, we implement a 2-minute timeout on NBD requests and
+    responses. If a request takes longer than that to complete,
+    the client is disconnected. This is useful to keep broken
+    clients from breaking migrations, among other things.
+
 listen
 ~~~~~~