Add "act" to preprocess_fn call in collector. (#801)

This allows, for instance, to change the action registered into the
buffer when the environment modify the action.

Useful in offline learning for instance, since the true actions are in a
dataset and the actions of the agent are ignored.

- [ ] I have marked all applicable categories:
    + [ ] exception-raising fix
    + [ ] algorithm implementation fix
    + [ ] documentation modification
    + [X] new feature
- [X ] I have reformatted the code using `make format` (**required**)
- [X] I have checked the code using `make commit-checks` (**required**)
- [] If applicable, I have mentioned the relevant/related issue(s)
- [X] If applicable, I have listed every items in this Pull Request
below

tianshou/data/collector.py

@@ -347,6 +347,7 @@ class Collector(object):
                         info=self.data.info,
                         policy=self.data.policy,
                         env_id=ready_env_ids,
+                        act=self.data.act,
                     )
                 )
 
@@ -626,6 +627,7 @@ class AsyncCollector(Collector):
                             truncated=self.data.truncated,
                             info=self.data.info,
                             env_id=ready_env_ids,
+                            act=self.data.act,
                         )
                     )
                 except TypeError:
@@ -636,6 +638,7 @@ class AsyncCollector(Collector):
                             done=self.data.done,
                             info=self.data.info,
                             env_id=ready_env_ids,
+                            act=self.data.act,
                         )
                     )