Closes #18690

Postgres and MySQL don't guarantee that pagination with `LIMIT` and
`OFFSET` will work if the ordering isn't unique. From the Postgres docs:

> When using `LIMIT`, it is important to use an `ORDER BY` clause that
> constrains the result rows into a unique order. Otherwise you will get
> an unpredictable subset of the query's rows

Before:

    [1] pry(main)> issues = 1.upto(Issue.count).map { |i| Issue.sort('priority').page(i).per(1).map(&:id) }.flatten
    [2] pry(main)> issues.count
    => 81
    [3] pry(main)> issues.uniq.count
    => 42

After:

    [1] pry(main)> issues = 1.upto(Issue.count).map { |i| Issue.sort('priority').page(i).per(1).map(&:id) }.flatten
    [2] pry(main)> issues.count
    => 81
    [3] pry(main)> issues.uniq.count
    => 81

[ci skip]
Signed-off-by: NRémy Coutable <remy@rymai.me>

Rails's form helpers use the `$attr_before_type_cast` method where
available, and this value only appears to be updated on assignment, not
when the object is mutated in some other way:

    [1] pry(main)> mr = MergeRequest.new
    => #<MergeRequest:0x007fcf28395d88 ...>
    [2] pry(main)> mr.description = 'foo'
    => "foo"
    [3] pry(main)> mr.description << ' bar'
    => "foo bar"
    [4] pry(main)> mr.description
    => "foo bar"
    [5] pry(main)> mr.description_before_type_cast
    => "foo"
    [6] pry(main)> mr.description += ' bar'
    => "foo bar bar"
    [7] pry(main)> mr.description_before_type_cast
    => "foo bar bar"

The method Banzai::Renderer.pre_process would always be called,
regardless of whether the Markdown to render was already cached or not.
In cache the document _was_ cached the output of the pre-processing
pipeline was ignored resulting in it doing nothing but wasting CPU
cycles.

This commit moves Banzai::Renderer.pre_process into
Banzai::Renderer.render_result so that it's _only_ used when needed.

    A lot of git operations were being repeated, for example, to build a url
    you would ask if the path was a Tree, which would call a recursive routine
    in Gitlab::Git::Tree#where, then ask if the path was a Blob, which would
    call a recursive routine at Gitlab::Git::Blob#find, making reference to
    the same git objects several times. Now we call Rugged::Tree#path, which
    allows us to determine the type of the path in one pass.

    Some other minor improvement added, like saving commonly used references
    instead of calculating them each time.

Previously we'd create a separate Metric instance for every method call
that would exceed the method call threshold. This is problematic because
it doesn't provide us with information to accurately get the _total_
execution time of a particular method. For example, if the method
"Foo#bar" was called 4 times with a runtime of ~10 milliseconds we'd end
up with 4 different Metric instances. If we were to then get the
average/95th percentile/etc of the timings this would be roughly 10
milliseconds. However, the _actual_ total time spent in this method
would be around 40 milliseconds.

To solve this problem we now create a single Metric instance per method.
This Metric instance contains the _total_ real/CPU time and the call
count for every instrumented method.

Updated shortcuts.png for docs

Updated CHANGELOG

Moved CHANGELOG entry

updated shortcut docs

- As todos are created/updated inside the TodoService
we repopulate the cache just there for both pending/done todos
- Todos as mark as done from the TodosController we update cache
there too
- All the added methods are kept in the User class for cohesion

Signed-off-by: NRémy Coutable <remy@rymai.me>