Make ShellcmdLrms handle app.requested_memory as total app memory.

`Application.requested_memory` is the *total* amount of memory requested by the application independently of the number of cores, as documented [in the `Application` docstring][1].

Now, different backends handle `Application.requested_memory` differently; see for example the behaviour of `Application.sbatch()` where `Application..requested_memory` is handled as a per-node requirement (see the documentation for the '--mem' option [in the SLURM man pages][2])

So the way to solve this problem is to make sure that `Application.requested_memory` is handled consistently for all backends, hence, change `ShellcmdLrms` accordingly.

[1]: https://github.com/uzh/gc3pie/blob/master/gc3libs/__init__.py#L765
[2]: https://slurm.schedmd.com/sbatch.html

gc3libs/backends/shellcmd.py

@@ -1357,7 +1357,7 @@ def submit_job(self, app):
         self.free_slots -= app.requested_cores
         self.user_run += 1
         if app.requested_memory:
-            self.available_memory -= (app.requested_memory * app.requested_cores)
+            self.available_memory -= app.requested_memory
         self._job_infos[pid] = {
             'requested_cores': app.requested_cores,
             'requested_memory': app.requested_memory,
@@ -1388,14 +1388,13 @@ def _check_app_requirements(self, app, update=True):
                     (self.name, self.max_cores))
 
         if app.requested_memory:
-            total_requested_memory = app.requested_cores * app.requested_memory
-            if self.available_memory < total_requested_memory:
+            if self.available_memory < app.requested_memory:
                 raise gc3libs.exceptions.MaximumCapacityReached(
                     "Resource {0} does not have enough available memory:"
                     " {1} requested total, but only {2} available."
                     .format(
                         self.name,
-                        total_requested_memory.to_str('%g%s', unit=Memory.MB),
+                        app.requested_memory.to_str('%g%s', unit=Memory.MB),
                         self.available_memory.to_str('%g%s', unit=Memory.MB))
             )
 
@@ -1629,7 +1628,7 @@ def _cleanup_terminating_task(self, app, pid, termstatus=None):
             self.free_slots += app.requested_cores
             self.user_run -= 1
             if app.requested_memory is not None:
-                self.available_memory += (app.requested_memory * app.requested_cores)
+                self.available_memory += app.requested_memory
         wrapper_filename = posixpath.join(
             app.execution.lrms_execdir,
             ShellcmdLrms.PRIVATE_DIR,

gc3libs/backends/tests/test_shellcmd.py

@@ -220,8 +220,7 @@ def test_resource_usage(self):
         cores_after = self.backend.free_slots
         mem_after = self.backend.available_memory
         assert cores_before == cores_after + 2
-        assert mem_before == mem_after + (app.requested_cores * app.requested_memory)
-
+        assert mem_before == mem_after + app.requested_memory
         self.run_until_terminating(app)
         assert self.backend.free_slots == cores_before
         assert self.backend.available_memory == mem_before