BUG/MEDIUM: streams: Remove SF_ADDR_SET if we're retrying due to L7 retry.

In do_l7_retry(), remove the SF_ADDR_SET flag. Otherwise,
assign_server_address() won't be called again, which means for 2.1 or 2.2,
we will always retry to connect to the server that just failed, and for 2.0,
that we will try to use to whatever the address is for the connection,
probably the last server used by that connection before it was pool_free()
and reallocated.

This should be backported to 2.1 and 2.0.

(cherry picked from commit 8cabc9783aaa8309f4ec6f7dd95cd74c61ed2e61)
Signed-off-by: Christopher Faulet <cfaulet@haproxy.com>
(cherry picked from commit 4d84a2d0d08eb4880cc8ea2c79924cebe9cc1fb9)
Signed-off-by: Willy Tarreau <w@1wt.eu>
diff --git a/src/proto_htx.c b/src/proto_htx.c
index ec08fb6..7b3464b 100644
--- a/src/proto_htx.c
+++ b/src/proto_htx.c
@@ -1413,6 +1413,7 @@
 	res->flags &= ~(CF_READ_ERROR | CF_READ_TIMEOUT | CF_SHUTR | CF_EOI | CF_READ_NULL | CF_SHUTR_NOW);
 	res->analysers = 0;
 	si->flags &= ~(SI_FL_ERR | SI_FL_EXP | SI_FL_RXBLK_SHUT);
+	s->flags &= ~SF_ADDR_SET;
 	stream_choose_redispatch(s);
 	si->exp = TICK_ETERNITY;
 	res->rex = TICK_ETERNITY;