The selenium crash when trying to create a response for a url
without a page class. It didn't seem a problem at first.
It could be related to the network of the computer or some
other configuration (python 2?).

Sometime the account list page is fine but anything after
that gets you an action needed.

When going to the history_link to get more info, we sometime
a page about fraud with tips to the user on how to stay safe.
A similar case was handled previously when creating new recipient.

We go on this page to retrieve the tpp_id.

I hardcode the message since it is not in the response anymore.

The default SCA methods were : sms, then app notification, then otp.
This patch allows the user to configure (in config file), using the preferred_sca
key, the list of authentication methods he wants to allow (for example, to have
SMS first, then otp, then App notification, etc...)

As for now, the following methods are supported by the module:
* notification (mobile application notification)
* sms (sms on your trusted mobile phone)
* otp (e-Code that must be generated on the mobile application)

device name (which appears in the website) can be customized with the configuration
file.

Sometimes different requests can temporarily return a HTTP 502 with a
maintenance message

The redirection from life insurance or bourse websites to the base account was
missing one step to complete and thus creating issues while trying to access to
the accounts since after failing the redirection we are logged out

The order summary document was properly handled only for the french website.

Sometimes we can face a page asking if we want to use another account or the
account currently used

new PER account to be typed starting with 'BNPP' instead of 'BNP'

It will be used to tell which transfer frequencies a module can handle.

Seems to be for "Comptes tiers" accounts.

We can end up on that page after a redirection,
so BrowserUnavailable is raised inside on_load.

user keep getting blocked account message on login

Sometimes, the life insurance account will be found in the same univers
as the parent account and sometimes it is not the case.
In addition to that, the iteration order of the univers for the
get_accounts was having an impact on whether the "parent" could be
filled or not for a life insurance account.

Now, we do the matching only after all the accounts have been retrieved.

Add a private _number field to non payment Accounts (loan,
lifeinsurance) to be consistent with the Account objects of payment
accounts.

This avoid a crash when we try to use find_object to match a life
insurance with its parent by comparing the _number value with all the
accounts.

It looks like that this error is returned for getComptesDebiteurs when
no there is no account that is allowed to do transfers.

Listing recipients or the recipient of an account might not be possible
when in unrelated univers (spaces).

For example, it can't be done when inside the "immobilier" loan space.

typing 'Plan Epargne En Actions' to PEA

There was some pretty old code related to categories.
Some code parsed double-spaces, which might date back to old cragr website.

An action needed can happen during the handle_sms,
but self.sms_form will have been re-initiated to None while the user resolve it,
and the OTP will already been submitted and accepted by the server.
So, to avoid running handle_sms a second time, we check if self.sms_form is present;
when not, we fall back to init_login, where the SCA won't be triggered.
This avoids a "'NoneType' object does not support item assignment"
error.

The warning message asks to use total_amount instead of price/income but
total_amount doesn't exist in Bill and it should be total_price
instead.

commit 8087d2f6dea023946894032a8fde7c1bac6438ce added the handling of
more error cases and a check for error messages at the `end_login()`
step to fix `FormNotFound` errors.

At this step, error messages and the message upon a succesful login are
located under the same XPath. So, we might raise false login errors.

To avoid false positives, we check whether we are logged in or not before handling
errors.

On the website, after login if the user uses strong authentication a
pop-in will show up. This pop-in's content is generated on the fly using
js and depends on a 'popin_dsp2' parameter code.

We use this parameter code to raise exceptions with the appropriate
corresponding messages.

we can be redirected on this page anywhere, not just at login

On BforBank website, when an user has several accounts, there is a
dropdown with all accounts available. The module iterate over this
dropdown and recover the value attribute of each <option> in order to build the
url that lead to saids accounts.

The previous regex that extract the value attribute was like this:

r'[^\d]' which allow to extract this value 123456789 and lead to build
this kind of url :

BASE_URL/espace-client/rib/123456789

This now lead to a 500 error and a lookalike unavailable/maintenance page.

The value attribute has obviously changed, and now look like this :

COMPTE_TITRE/yLsl6hUcNO_FbC7i9i0fUkFdESgVzGF2nwVuHfQhs08=

PEA/Alql9hUcDO_FbL4i9i4fUoPdJSgQzHF2nwVuHfQhs08=

etc...

r'/(.+)' allow to extract the value following the slash and build
this kind of url:

BASE_URL/espace-client/rib/wk4C9nIlhkw8-DGPATODM4QuZgPTf5-RQb1ftWHpoqA=

In some scenarios, relogin are triggered. If it happens inside
`handle_otp`, the otp value stored in config will not be cleared.

The consequence is that we will try to validate the otp once again, even
though we might not even be on the right page.

To fix that we reset the otp value to its default value manually.

Handle more login error messages, updated XPath for errors in
LoginPage's `on_load()` method.

Additionaly, added login errors check in `end_login()` to avoid
FormNotFound exception due to a blocked account access.