This adds a new method that capsulates the access check that has to be
done to decide if an admin plugin's page should be shown to the user.
The default implementation is the same as before, relying only on the
forAdminOnly() method and the users' isadmin or ismanager status.

Admin plugins themselves can override the method to do additional
checks. In this patch, I added that to the usermanager plugin which will
only return true if the current auth backend can list users.

However the real idea behind this change is that the new method emits a
new event called ADMINPLUGIN_ACCESS_CHECK which would allow plugins to
overwrite it. This way it could be possible to give certain user groups
access to certain admin plugins without giving them admin or manager
permissions.

Note: this does not change how the "Admin" link is shown, it still
depends on ismanager or isadmin. A plugin as mentioned above would need
to influence the display via the MENU_ITEMS_ASSEMBLY event.

Note: this only covers the basic access check. Admin plugins may need
further adjustments for access to other parts of the plugin (like AJAX
components). An additional commit will update this for the bundled
plugins.

Specifically, this pull request checks if the current query is already a
valid pageid and then return that as suggested pageid.

Also, tests are added for this method.

Fixes #2355

PS: We may want to somehow better educate/nudge users to use the better
way of creating pages by creating links on existing pages.

Also, for consistency this initializes also $num with 0, while ensuring
that the first FT_SNIPPET_NUMBER results with a $cnt>0 have a snippet.

Punctuation is more natural, makes sense with CSS switched off
and makes screenreaders add a little pause.

In #2286 the 2nd level heading in the intro was removed, so the h3
should be promoted to h2.

This implements some of the changes discussed in #2314:

1. last mod and number of hits are on the same line and always shown
2. There is a class around the last mod string so it can be
hidden/replaced by an icon
3. All items in the resultBody returned by the event are wrapped in
`<dd>` tags to ensure validity. They get their array key as class
4. There is still a wrapper div around each result for styling purposes

Recent changes to the search results page added various divs
to the pre-existing description list, which is invalid.
This fixes that by a) removing a couple of unnecessary elements
and b) interpret the 'last modified' line as another 'dd'.

The advanced search options list links with an invalid empty target
attribute. This fixes that by skipping empty parameters.

This was unintentionally removed in
4eab6f7c

It also should now be clearer what the code does.

This fixes #2305

They are now named min and max respectively, which is hopefully less
cryptic.

This is useful for anonymously useable forms, like the search.

As suggested by @splitbrain SearchState now behaves simliar to the PSR-7
message interfaces.

Implements the functionality requested in #1124

This removes a lot of reudundant CSS classes in favor of cascading
selectors. Styling isn't done, yet this is just a first step.

Still to do:

* aria labels
* styling the toggle mechanism
* RTL styling
* making sure everything displays correct when no JS is available
* translations may be missing
* refactoring the rest of search.less to make use of LESS features

This also renames the event for the search form on the search page.

The scheme for naming events that provide the new Form-class as $data is
FORM_<formname>_OUTPUT

If the entered query does not cleanly fall into the area of the given
options, show the current behavior as "custom".

This gives create control over when a query is too complex for tool
support.

This functionality is inspired by what other large search engines are
doing.

Building the new search links is complex and we're going to add another
parameter with the new sorting by mtime. Extracting a new class seems
like the cleanest way to handle that increasing complexity.

This is the more appropriate place for that functionality, because now
it happens inside the default function for the respective pagesearch and
pagelookup events and can be properly handled by plugins.

This way we do not loose the context of the current page. Further, the
new id generated from the query before wasn't really that useful (see
issue #1124 ). And we can still generate a meaningful link "create a
page for your query", if that is considered useful.

Further redirects exist for the old urls, so bookmarked searches etc.
should mostly keep working.

This allows filtering of results by the last modified time. It still
needs a custom date entry option, highlighting of the currently selected
option (if any) and a better place where the filtering happens.

This simplifies many aspects. However, it still needs much better
styling.

This adds two new config options:

`search_limit_to_first_ns`:
Limit the search to the current X namespaces. When a search is executed
from a page within a deeper namespace, the first X namespaces will be
added as filter.
Possible use case could be with language namespaces to ensure that the
default search is initially within the current language.

`search_default_fragment_behaviour`:
Option to specify the default fragment search behavior