fix django2.0

[opapy]

Failed set_many with django2.0.

>>> from django.core.cache import cache
>>> cache.set_many({'aa': 1})
Traceback (most recent call last):
  File "/usr/lib/python3.6/code.py", line 91, in runcode
    exec(code, self.locals)
  File "<console>", line 1, in <module>
  File "/usr/local/lib/python3.6/dist-packages/ddtrace/contrib/django/cache.py", line 65, in wrapped
    return method(*args, **kwargs)
  File "/usr/local/lib/python3.6/dist-packages/django_elastipymemcache/memcached.py", line 43, in wrapper
    return f(self, *args, **kwds)
  File "/usr/local/lib/python3.6/dist-packages/django_elastipymemcache/memcached.py", line 135, in set_many
    return super(ElastiPyMemCache, self).set_many(*args, **kwargs)
  File "/usr/local/lib/python3.6/dist-packages/django/core/cache/backends/memcached.py", line 144, in set_many
    return [original_keys[k] for k in failed_keys]
TypeError: 'bool' object is not iterable

Release Note: https://docs.djangoproject.com/en/2.0/releases/2.0/#cache
Code: https://github.com/django/django/blob/338f741c5eb6b91118f6a6b7c34b5e9b47a5661d/django/core/cache/backends/memcached.py#L131-L139

Since django 2.0, the return value of set_many now expects a list of keys of failed objects in store.

It should be implemented at the cache backend, but python-memcached and pylibmc return list objects, so I tried it accordingly.

[jogo]

Looks good, although this does change the behavior of the API so we should make sure to document that in the changelog. but the change behavior is fairly minor and looks reasonable. This doesn't seem to fully implement django's expected behavior of returning a list of failed keys, is that expected?

[opapy]

hi @jogo
Sure, this implementation returns a list of keys failed to store in HashClient only.

PooledClient and Client always return empty list.
Client and PooledClient is always True, if it fails it returns an exception, so I thought whether this implementation would be a problem.

[jogo]

so this means, if any of these keys fail, we mark all as failed?

[opapy]

thanks!
That's true. I will fix it.

[Surgo]

@jogo any update on this issue?
This PR is blocker of django 2.0 migrations...

[jogo]

looks good, one comment

[jogo]

doesn't this now return a list of failed sets.

[opapy]

thx! i fixed it.

[opapy]

pls restart job
https://travis-ci.org/pinterest/pymemcache/jobs/416230799

[nichochar]

Job has been restarted

[jogo]

Thanks looks good. Will work on releasing this sometime next week as a major version bump.

[opapy]

@jogo
thunks for review!

[Surgo]

@jogo When release it?

[jparise]

@Surgo I think we're on track for a 1.5.0 release pretty soon. I'd like to see #184 reviewed and merged first, and then we can start preparing a new release.