Flexible node allocation fails with local scheduler

[vkarak]

With #1074 we allowed local scheduler to support flexible allocation, but this is actually crashing now:

diff --git a/unittests/resources/checks/hellocheck.py b/unittests/resources/checks/hellocheck.py
index 62abd1d..7956eb8 100644
--- a/unittests/resources/checks/hellocheck.py
+++ b/unittests/resources/checks/hellocheck.py
@@ -9,6 +9,7 @@ class HelloTest(rfm.RegressionTest):
         self.descr = 'C Hello World test'

         # All available systems are supported
+        self.num_tasks = 0
         self.valid_systems = ['*']
         self.valid_prog_environs = ['*']
         self.sourcepath = 'hello.c'

Command line: ./bin/reframe -c unittests/resources/checks/hellocheck.py -r
Reframe version: 2.21-dev2
Launched by user: karakasv
Launched on host: dhcp-133-23.cscs.ch
Reframe paths
=============
    Check prefix      :
    Check search path : 'unittests/resources/checks/hellocheck.py'
    Stage dir prefix     : /Users/karakasv/Repositories/reframe/stage/
    Output dir prefix    : /Users/karakasv/Repositories/reframe/output/
    Perf. logging prefix : /Users/karakasv/Repositories/reframe/perflogs
[==========] Running 1 check(s)
[==========] Started on Thu Dec  5 18:45:48 2019

[----------] started processing hellocheck (C Hello World test)
[ RUN      ] hellocheck on generic:login using builtin-gcc
[     FAIL ] hellocheck on generic:login using builtin-gcc
[----------] finished processing hellocheck (C Hello World test)

[  FAILED  ] Ran 1 test case(s) from 1 check(s) (1 failure(s))
[==========] Finished on Thu Dec  5 18:45:48 2019

==============================================================================
SUMMARY OF FAILURES
------------------------------------------------------------------------------
FAILURE INFO for hellocheck
  * System partition: generic:login
  * Environment: builtin-gcc
  * Stage directory: /Users/karakasv/Repositories/reframe/stage/generic/login/builtin-gcc/hellocheck
  * Node list: <None>
  * Job type: local (id=None)
  * Maintainers: ['VK']
  * Failing phase: run
  * Reason: unexpected error: 'str' object has no attribute 'is_available'
Traceback (most recent call last):
  File "/Users/karakasv/Repositories/reframe/reframe/frontend/executors/__init__.py", line 163, in _safe_call
    return fn(*args, **kwargs)
  File "/Users/karakasv/Repositories/reframe/reframe/core/pipeline.py", line 77, in _fn
    func(obj, *args, **kwargs)
  File "/Users/karakasv/Repositories/reframe/reframe/core/pipeline.py", line 1171, in run
    self._job.prepare(commands, environs)
  File "/Users/karakasv/Repositories/reframe/reframe/core/schedulers/__init__.py", line 270, in prepare
    guessed_num_tasks = self.guess_num_tasks()
  File "/Users/karakasv/Repositories/reframe/reframe/core/schedulers/__init__.py", line 308, in guess_num_tasks
    available_nodes = {n for n in available_nodes
  File "/Users/karakasv/Repositories/reframe/reframe/core/schedulers/__init__.py", line 309, in <setcomp>
    if n.is_available()}
AttributeError: 'str' object has no attribute 'is_available'

------------------------------------------------------------------------------

The problem is that we assume (wrongly) that the nodes returned are of type SlurmNode, whereas the local scheduler returns strings. We should create an API there as well.