Improve security release documentation

I think the MailChimp step is not needed anymore because we automated it.

See merge request !1271

Properly fix wiki restore

My previous patch didn't fix the restore issue (!247).

ProjectWiki.new() creates a new wiki git repository, so any tries to bare clone a bundle fail. With this patch we remove the newly created wiki.git before restoring from the backup bundle.

\cc @jacobvosmaer @dzaporozhets are you ok with this solution/hack?

Relevant issues: #845

See merge request !248

ProjectWiki.new() creates a new wiki git repository, so any tries
to bare clone a bundle fail. With this patch we remove the newly
created wiki.git before restoring from the backup bundle.

Signed-off-by: Dmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

Conflicts:
	CONTRIBUTING.md

Signed-off-by: Dmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

Reactivate highlight.js language autodetection

### What does this MR do?

This commit partly reverts commit 0d13abb1 to reactivate highlight.js language autodetection.

It does NOT revert the part that adds the `nohighlight` CSS-class for a predefined set of filenames (which was indeed the intended purpose of commit 0d13abb1).

### Are there points in the code the reviewer needs to double check?

Not particularly, the only effect of this blob_helper is to add a CSS class to the `<pre>` container of the blob, to control the way highlight.js performs its syntax highlighting. (see where this helper is called at https://gitlab.com/gitlab-org/gitlab-ce/blob/master/app/views/shared/_file_hljs.html.haml#L12).

However this merge request DOES change the way the syntax highlighting is done (by bringing it back to the situation before commit 0d13abb1 where highlight.js autodetected the language itself). So we can expect some new kind of syntax highlighting problems. However in this case, the issues should be forwarded to the upstream project (https://github.com/isagalaev/highlight.js).

### Why was this MR needed?

The main motivation behind this partial revert is that there is no point (IMHO) in using an "intelligent" syntax highlighter which is specifically designed to autodetect languages if it is to dictate (via the CSS class) the language to use solely based on the filename extension.

### What are the relevant issue numbers / Feature requests?

Issue #665 "HLJS language autodetection / syntax highlighting fails "

### Screenshots (If appropriate)

See screenshots in issue #665

See merge request !232

Fixed an alignment issue, fixes #778

This merge request fixes issue #778. I have figured out the problem already here: https://gitlab.com/gitlab-org/gitlab-ce/issues/778#note_366548

See merge request !242

Fix wiki restore

Seems 4a5044e3 only fixed git repos restore. This MR provides the same functionality that was introduced in the aforementioned commit, but for wikis.

See merge request !247

Make clear that the upgrader script does not update gitlab-shell

There are many cases where people end up with a broken update because they don't update gitlab-shell.

For example: https://twitter.com/hgomez/status/537902038907363330

\cc @sytses

See merge request !246

Create project & repository at the same time

* block project creation if repository cant be created
* show proper error message about failed to create repository
* dont use ajax for project creation any more - its fast enough
* if project repository does not exist - propose to create one or import
* move import functionality to separate resource

Fixes #1768

See merge request !1278

Signed-off-by: Dmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

Signed-off-by: Dmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

Signed-off-by: Dmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

Signed-off-by: Dmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

Signed-off-by: Dmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

Sidekiq memory killer

This change adds an optional Sidekiq memory killer middleware that
will perform a graceful shutdown when the Sidekiq process is using
too much memory. When combined with a process supervision daemon
(such as Runit in omnibus-gitlab) this should provide Good Enough
handling of the Sidekiq memory leaks we are seeing.

The memory killer is off by default.

See merge request !1276

When enabled, this middleware allows Sidekiq to detect that its RSS has
exceeded a maximum value, triggering a graceful shutdown. This
middleware should be combined with external process supervision that
will restart Sidekiq after the graceful shutdown, such as Runit.

Add Dockerfile to build an Omnibus GitLab image

See merge request !244

Office analogy

introduced office analogy to architecture.md

See merge request !1275

add details on protected branches

Add some details from https://about.gitlab.com/2014/11/26/keeping-your-code-protected/:

Who can:
* Force push to non-protected branches
* Force push to protected branches 
* Remove protected branches

Select a tag

Related to #1603.

Prevent users looking to install GitLab from source from installing a release candidate by mistake.

See merge request !1274

Update time zone rake task for production. Resolves #8387

Conflicts:
	GITLAB_SHELL_VERSION

Security release ownership

We saw the following problems around the 7.4.4 release:

- the fix got merged into master while it should have been held back
- there was a security fix but no release manager to release it

With this change we want to improve our process and documentation
to prevent situations like this in the future.

See merge request !1273