Do not require positional args for #authenticate

This API is a little bit confusing, IMO.  But it does preserve backward
compatibility, while allowing authenticators that don't allow positional
parameters to work without crashing.  But, authenticators that require
only one parameter—or more than three—will still be inaccessible.

Author: nevans

lib/net/smtp.rb

@@ -862,6 +862,8 @@ def open_message_stream(from_addr, *to_addrs, &block)   # :yield: stream
     DEFAULT_AUTH_TYPE = :plain
 
     # call-seq:
+    #   authenticate(type: DEFAULT_AUTH_TYPE, **, &)
+    #   authenticate(type = DEFAULT_AUTH_TYPE, **, &)
     #   authenticate(username, secret, type: DEFAULT_AUTH_TYPE, **, &)
     #   authenticate(username, secret, type = DEFAULT_AUTH_TYPE, **, &)
     #
@@ -872,11 +874,17 @@ def open_message_stream(from_addr, *to_addrs, &block)   # :yield: stream
     # All arguments—other than +type+—are forwarded to the authenticator.
     # Different authenticators may interpret the +username+ and +secret+
     # arguments differently.
-    def authenticate(user, secret, typearg = nil, type: nil, **kwargs, &block)
+    def authenticate(*args, type: nil, **kwargs, &block)
+      case args.length
+      when 1, 3 then typearg = args.pop
+      when (4..)
+        raise ArgumentError, "wrong number of arguments " \
+                             "(given %d, expected 0..3)" % [args.length]
+      end
       authtype = type || typearg || DEFAULT_AUTH_TYPE
-      check_auth_args(authtype, user, secret, **kwargs)
+      check_auth_args(authtype, *args, **kwargs)
       authenticator = Authenticator.auth_class(authtype).new(self)
-      critical { authenticator.auth(user, secret, **kwargs) }
+      critical { authenticator.auth(*args, **kwargs) }
     end
 
     private