Convert clear_ready to use triggers (#9)

This prevents a window where other handlers might run based on the layer
flags after the interface is disconnected but before the clear_ready
handler has a chance to run.

Part of the fix for LP: #1844103

reactive/vault_kv.py

@@ -1,11 +1,17 @@
 from charmhelpers.core import hookenv, host
 from charms.reactive import when_all, when_not, set_flag, clear_flag
-from charms.reactive import endpoint_from_flag
+from charms.reactive import endpoint_from_flag, register_trigger
 from charms.reactive import data_changed
 
 from charms.layer import vault_kv
 
 
+register_trigger(when_not='vault-kv.connected',
+                 clear_flag='layer.vault-kv.ready')
+register_trigger(when_not='vault-kv.connected',
+                 clear_flag='layer.vault-kv.requested')
+
+
 @when_all('vault-kv.connected')
 @when_not('layer.vault-kv.requested')
 def request_vault_access():
@@ -37,12 +43,6 @@ def check_config_changed():
             set_flag('layer.vault-kv.config.changed')
 
 
-@when_not('vault-kv.connected')
-def clear_ready():
-    clear_flag('layer.vault-kv.ready')
-    clear_flag('layer.vault-kv.requested')
-
-
 def manage_app_kv_flags():
     try:
         app_kv = vault_kv.VaultAppKV()