Move 'mark_read' call out of 'get_context_data' and into 'get'.

[skolsuper]

Two reasons for moving the method call:

1. I wouldn't expect 'get_context_data' to have side effects.
2. It ensures that the view returns without error before marking the topic read

This PR also modifies an inefficient query in mark_read that checks for unread topics in forum.

[DylannCordel]

LGTM but If I may bring my 2cts:

• IMHO, mark_read call should not be conditioned. Your condition should be in the mark_read method (with a direct return) to allow developers to easily override this method and do special stuff if user is not authenticated (for example, mark ad read via Cookies ?)
• you should add an atomic transaction for all the mark_read method. I think we could have some side effect if someone post a new post in the same time (but a microsecond after), and the sever run this request before you delete Trackers : you'll delete a newly created tracker.
• I think that disabling a test is never a good idea. With a DB that does not support microsecond, you don't test parts of code that will be ran with this type of DB. So you hide some potential bugs. Maybe you could add a time.sleep(1) if DB does not support microsecond between the post creation and the post access ?

[DylannCordel]

Forget the third point : I'm working on a PR to remove those @skipUnlessDBFeature. There is still a test which always fails with MySQL, even with sleep calls. I think there is a bug in pybbm read tracking system with MySQL. If you want to have a look on the WIP : webu/pybbm@ec7d7d808be5237b6d550e5c01ebba5274d50194

[DylannCordel]

@skolsuper : see #258 for the third point.

[skolsuper]

Agree with points 1 and 2 but don't have time to fix. I have given push access to my fork so you can touch up this branch if you think it's worthwhile.

[DylannCordel]

@skolsuper I rebase the PR and add points 1 and 2. Thanks for the access.

[lampslave]

Better late than never :) Thank you.

[coveralls]

Changes Unknown when pulling 25b761ca6f3706f786631457cb8200a5124b4f31 on skolsuper:mark_read into on hovel:master.