See https://github.com/owncloud/mirall/commit/eb7074e9f088214648183a45c2ce806f65a8bf3c for discussion

Should fix #1502 and #1524
and #1459 and #1521

Also only run the propfind job if the credidentials are readon, and fetch them
otherwise

This reverts commit 9eb3452a.

We need that or wrong password can never be changed

This should fix bug #1491

Also set the job as parent to get proper deletion of the
settings object.

That confused win32, since this flag was true very often.

Fixes #1491

If we never fetched the credidentials, the _user and _password will be empty,
we should therefore not try to save them because it would just erase them.

Version 1.5.2 always try to fetch the creds at startup.  But now we only
do it if we detect an internet connection

Sometimes 'success' is false without apparent reason.
Just remove the message

They will be fetched by the connection validator after doing
the status.php check

This reverts commit 86073001.

This approach is too dangerous for 1.5.2.

If we keep the cookies, next time we sign in, the entered password
does not matter as the server still think our session is logged in.

It is likely to re-enter if there is two jobs that asks for a password.

Example:
 1. log out
 2. restart the application
 3. enter a wrong password
 4. enter a wrong password again a few times
 5. enter the correct password
 6. it should must not prompt for the password again.

Because of the re-entrency, it was still prompting for the password several
times after the right password had been entered

If a QNAM job (e.g. Quota or ETag job) gets redirected,
we'll invalidate and then prompt to re-fetch the credentials
from the user.
(The keychain credentials will be wrong so they get deleted)

We need to reset the credential type when
returning to the first page from later steps.
Otherwise, unfinished credential instances
and their NAM will try to fetch data, which
will fail with weird symptoms.