This caches the output of Repository#exists? in Redis while making sure
it's flushed properly when creating new repositories, deleting them,
etc.

For the ProjectWiki tests to work I had to make ProjectWiki#create_repo!
public as testing private methods in RSpec is a bit of a pain.

The avatar logic has been moved from Project to Repository as this makes
caching easier. The logic itself in turn has been changed so that the
logo file names are cached in Redis. This cache is flushed upon pushing
a commit but _only_ if:

1. The commit was pushed to the default branch
2. The commit actually changes any of the logo files

If no branch or commit is given the cache is flushed anyway, this
ensures that calling Repository#expire_cache without any arguments still
flushes the avatar cache (e.g. this is used when removing a project).

Fixes gitlab-org/gitlab-ce#14363

The queries that are UNION'd together don't need any eager loading
(since we really only use the resulting SQL instead of having
ActiveRecord actually run the queries). By dropping any eager loaded
associations queries such as the following work instead of producing a
SQL error:

    Project.all.includes(:namespace).search('foo')

This solves the problem with caching the nil value with instance variable.
Without this the every time we ask for external_issue_tracker we built AR and potentially do SQL query

We can just use "arel_table" in these cases instead of
"SomeClass.arel_table".

Similar to the changes made to Project.search the method
Project.search_by_title now also uses Arel so it can automatically use
ILIKE/LIKE instead of the lower() function.

This chance is broken up in two steps:

1. Use ILIKE on PostgreSQL and LIKE on MySQL, instead of using
   "WHERE lower(x) LIKE lower(y)" as ILIKE is significantly faster than
   using lower(). In many cases the use of lower() will force a slow
   sequence scan.

2. Instead of using 1 query that searches both projects and namespaces
   using a JOIN we're using 2 separate queries that are UNION'd
   together. Using a JOIN would force a slow sequence scan, using a
   UNION avoids this.

This method now uses Arel as Arel automatically uses ILIKE on PostgreSQL
and LIKE on MySQL, removing the need to handle this manually.

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

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

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

The method Project#ci_services would load all services into memory
(including _all_ their columns) and then use Enumerable#select to reduce
the list. Project#ci_service in turn would further reduce this list down
to just 1 Service instance.

Instead of doing all this in Ruby we can just offload the work to the
database, reducing the amount of time spent in these methods. These
changes reduce the time of the first call to Project#ci_services from
around 240 ms to around 10 ms, though the final timings will vary based
on database load. Because Project#ci_service is memoized there's no
further overhead introduced by using a database query.

Fixes gitlab-org/gitlab-ce#14186

Prevent Groups to have smaller visibility than projects
Add default_group_visibility_level to configuration
Code improvements

removes automatic setting of main_language to project for it to set a main language you have now to make a commit to the project

Make methods return ActiveRecord Relations instead of Arrays.