Remove the filename argument from Content-Disposition header

This MR removes the filename argument from the `Content-Disposition` header to avoid RFC 5987 and RFC 6266 encoding issues. Some browsers (e.g. Internet Explorer) do not properly decode a Unicode string properly, and this can lead to odd filenames in the raw file download. This change allows the browser to determine the filename based on the URL.

For example, if I have a file called `한글한글.pptx` and click to download it with the "Raw" button in the Files section, IE11 will ask:

![image](https://gitlab.com/stanhu/gitlab-ce/uploads/fdb688282c0d4564872deadb44c58b2c/image.png)

If you look at the `Content-Disposition` field, you see this:

![image](https://gitlab.com/stanhu/gitlab-ce/uploads/2b342b83ae1ec61fd31937163ace8ec5/image.png)

Chrome, Firefox, and Safari seem to be able to handle UTF-8 encoded filenames, even though this is not standard.

See: http://greenbytes.de/tech/tc2231/

Closes https://github.com/gitlabhq/gitlabhq/issues/9595

Closes https://gitlab.com/gitlab-org/gitlab-ce/issues/1829

I've also submitted a pull request to Rails to support RFC 6266: https://github.com/rails/rails/pull/21461

See merge request !1235

Drone CI service

https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/1184#note_1980652

/cc @DouweM

See merge request !1215

RFC 5987 and RFC 6266 encoding issues. This change allows the browser
to determine the filename based on the URL.

See: http://greenbytes.de/tech/tc2231/

Closes https://github.com/gitlabhq/gitlabhq/issues/9595

Closes https://gitlab.com/gitlab-org/gitlab-ce/issues/1829

Check if database connection exists when loading application settings

!1217 broke omnibus-gitlab nightly builds. In omnibus-gitlab we precompile assets without a running database.

See merge request !1230

Decouple Gitlab::Markdown from the GitlabMarkdownHelper

This module is now the sole source of knowledge for *how* we render
Markdown (and GFM).

See merge request !1223

Import all Bitbucket issues and sort issues by creation date

Previously only the first 15 Bitbucket issues would be imported. Other issues in the API return call also calls import errors.

Tested by importing 400+ issues from this repo: https://bitbucket.org/ned/coveragepy

API reference here: https://bitbucket.org/site/master/issues/3571/api-issues-sorting-bb-3518

Closes https://github.com/gitlabhq/gitlabhq/issues/9519

See merge request !1226

Also fix a number of issues where author/reporter names were not available.

API reference here: https://bitbucket.org/site/master/issues/3571/api-issues-sorting-bb-3518

Closes https://github.com/gitlabhq/gitlabhq/issues/9519

Replace grack with gitlab-git-http-server

See merge request !1229

Rename profile settings navigation to "Profile Settings" 

Fixes #2555 

See merge request !1926

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

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

Update Racc to 1.4.12

The primary reason for this is that this would let me debug #2343 using Rubinius' `-Xic.debug` option.

See merge request !1201

Added new commands to GitLab Basics docs



See merge request !1924

Expire cache when merge request source branch was removed

Re-implement !1160 for 8.0. Also we have tests for MergeWorker now! Yay ✨Signed-off-by: NDmitriy Zaporozhets <dmitriy.zaporozhets@gmail.com>

See merge request !1224

Retrieving oauth token with LDAP credentials

https://dev.gitlab.org/gitlab/gitlabhq/issues/2529

See merge request !1220

This adds some duplication, but this helper is temporary.

There wasn't really a reason to have them as a constant, and we were
getting "already defined" warnings which are always annoying.

Some usages of the `gfm` helper depend on the text not being rendered as
standard Markdown. Ugh.

We need to send `path`, not `requested_path`.

Prevent anchors from being hidden by header upon a new page load

Browsers change the scroll position after the page is loaded, and the current 1 ms
delay wasn't long enough for the browser to jump to the anchor. Even then, it appears
another 100ms delay is necessary to adjust the scroll properly:

http://stackoverflow.com/questions/19057731/scrolltop-on-page-reload-doesnt-work-possible-script-conflict

Add a little more of an offset to see a few lines before the anchor.

* Closes #1614
* Closes #1024

See merge request !1225

See https://gitlab.com/gitlab-org/gitlab-ce/commit/4a203aab57c69ea402939cf8fc423a0e5aaa1cab

Don't notify users without access to the project when they are (accidentally) mentioned in a note.

Addresses #2366.

See merge request !1216

Browsers change the scroll position after the page is loaded, and the current 1 ms
delay wasn't long enough for the browser to jump to the anchor. Even then, it appears
another 100ms delay is necessary to adjust the scroll properly:

http://stackoverflow.com/questions/19057731/scrolltop-on-page-reload-doesnt-work-possible-script-conflict

Add a little more of an offset to see a few lines before the anchor.

Closes #1614
Closes #1024