The read_progress_bar method is incredibly not efficient. It is also
wrong in other ways (for example, if 2 groups have same labels, there is
a collision). But let us talk about efficiency: it iterates on each
records on the domain, then, it computes some information. It is really
worse for selection fields: it performs a fields_get for each iteration.
This commit improves the situation by doing the fields_get only once.
Note that this is not the end of the story: this only has a small effect on
the performance of this method. It looks like there is a deeper issue in
some cases (kanban view for subscription, field 'activity_state')