[spirica] add currency field on the account, changes xpath on the table to make it adaptable to different currencies

The website displays a restriction on stock market functions.

Closes: 595486@redmine
Closes: 628958@redmine

It now looks for a message in the table.

Closes: 625530@redmine
Closes: 626587@redmine
Closes: 625354@redmine
Closes: 625683@redmine

When website return a 500 error code self.page is not updated
so we won't iter history on self.page which is the same handled page
but formatted differently if the first location as failed

Closes: 624005@redmine

Closes: 630268@redmine

Closes: 4956@zendesk

get page with partialHTMLpage

Closes: 631857@remdine

don't crash handling non-mandatory fields
instead we fill them with the new FetchError EmptyField

Closes: 622718@redmine 595926@redmine

The website slightly changed its textual content and nothing is matched
by the xpath. Consequently a wrong input is passed to Decimal (an empty string) by the
CleanText filter.

Closes: 549715@redmine

For some accounts, it seems like there is an prior optional page in order
to visit the Espace bourse, that contains a form to submit.

Closes: 4668@zendesk
Closes: 573510@redmine

Change description parameter into message parameter
to show "wrong code" error message to users

The $'xxx' syntax is not standard shell syntax so it wasn't interpreted
correctly, causing pyflakes to spit a lot of errors.
This is after porting shebang to plain sh.

- use env(1) to found bash on the system
- remove bash dependency on few scripts

Also, fix location fetching.

64cad1db was supposed to set fields to
NotAvailable when the backend supports filling but doesn't fill a field.

Unwanted effect was that if backend doesn't support filling at all, an
incomplete field would be set to NotAvailable.

And store it if retrieved.

If vignette is installed and an XDG thumbnail matching an object's URL
and datetime, don't call fillobj for retrieving thumbnail icon.

Some modules can return a lot of items. Instead of consuming all results
immediately, results can now be fetched on demand. That avoids having to
set a low limit to number of results.
This is done by not closing the iterator iter_gallery_images etc.
The app uses a Queue to pause/resume iteration.

Also, sometimes it's useful to do the same search, but some results are
unwanted, so we may want to hide them permanently, to avoid fetching their
data and thumbnail next time the iterator reaches them.
Items can be bookmarked or ignored.

Store an intermediate Item class in QModelIndex::internalPointer. This
will make it easier to change a weboob object when a fillobj method
returns a new object instead of filling in-place.

The RSS class changed, updating it fixes the podcast feature
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@no-log.org>

Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@no-log.org>

TODO CapMessagesPost