drupal:drush execute the same command on all servers.

[theist]

Due the nature of capistrano's ask method and the parallelism of the roles method the value of fetch(:drupal_command) is undefined for every run of the roles loop

This commit forces the evaluation of :drupal_command before enter the roles loop, efectively launching the same command on each host in parallel

Whitout this commit:

22:44 $ cap production drupal:drush
Using airbrussh format.
Verbose output is being written to log/capistrano.log.
Please enter drush_command (Drush command you want to run (eg. 'cache-clear css-js'). Type 'help' to have a list of avaible drush commands.): Please enter drush_command (Drush command you want to run (eg. 'cache-clear css-js'). Type 'help' to have a list of avaible drush commands.): cc all
00:00 drupal:drush
      01 /var/www/test/shared/vendor/bin/drush cc all 
    ✔ 01 ubuntu@server1 5.238s
help
      02 /var/www/test/shared/vendor/bin/drush help
    ✔ 02 ubuntu@server2 0.657s

As you see the "promp" question for ask appears twice, then capistrano executes in one server and then appears to be hang. But it is waiting for the input of "another" ask

with this commit

22:47 $ cap production drupal:drush
Using airbrussh format.
Verbose output is being written to log/capistrano.log.
Please enter drush_command (Drush command you want to run (eg. 'cache-clear css-js'). Type 'help' to have a list of avaible drush commands.): cc all 
00:00 drupal:drush
      01 /var/www/test/shared/vendor/bin/drush cc all 
    ✔ 01 ubuntu@server1 5.043s
    ✔ 01 ubuntu@server2 5.305s

Assignation forces preevaluation of the ask before launch commands.