Add dynamic counts to search filters

[jorg-vr]

This pull request contains a refactor of search filters. It adds dynamic counts of the expected results to all search filters. When already partially filtered, it also limits the other filters to only show relevant options.

The initial goal was to get rid of some old code that used js functions as part of 'filterCollection' objects to extract the color and paramVal from a filter option. This made most related code quite ugly and there was no longer any need for it.

So with cc610d6 and 810fa47 I made filterCollection a pure map of data objects.

With these changes it became a lot simpler to dynamically update filters.
I tried to write generic concerns that automatically define a dynamic option count when defining a scope. This way the end result should be less specific code.

In the most basic case writing a filter for a given field of a model is as simple as writing filterable_by :field in the model and writing has_filter :field in the controller. In the controller, the @filters variable can then be set by calling filters(elements) on the unfiltered list of elements. @filters is then used to update the shown filter options and counts.
Common cases, such as enums and foreign keys can be filtered using filterable_by :field, is_enum: true and filterable_by :field, model: Model respectively.

Naturally there are always some more complex cases, in which case a custom name_hash function can be written or even a full custom count function. See activity.rb for some examples of this.

The most complex case were course_labels as these required the course as extra argument. For now these are generalized in their own specific reusable function.

[chvp]

(There are still some test and linting issues here.)

[jorg-vr]

Maybe I overreacted.

The newly added count queries have a similar speed as the existing count query that already ran for pagination.
For most pages it doesn't have a big impact. Most pages only have one or two filters and this results into at most 100ms extra response time on naos, but often a lot less.

I have already removed the counts for the submission pages, as it was unbearably slow there (but these pages where already way to slow on naos anyway).
For reference the pagination count takes 15.8 seconds out of the total 16.7 seconds on the submissions page. Adding up to two extra count queries to this page tripled response time.
Side note: removing the page count on this page would greatly speed up the page.

The activities page is an edge case. With 7 filters the increase in query time can be up to 300ms, which is about 20% of the total time required to render the page. This makes interactivity on naos feel slightly suboptimal.
But our production server is always a bit faster, and the UX improvement is greatest for activity selection, so it might be worth a 20% increase in loading time.
This page also has room for speed ups in other places, such as removing the need for n+1 queries for popularity and activity status.

[chvp]

This also seems weird to me. It should be possible to delegate to the controller instance instead of the class, which would be better, IMO.

[chvp]

Can we add an extra method to the concern that handles cases like this (and popularity below). It also occurs in submission, IIUC.

[jorg-vr]

I have now added this extra method:

def filter_options_for(name, &block)
  define_singleton_method("#{name}_filter_options", &block)
end

Which does make it a bit more readable. But except for hiding the define_singleton_method it doesn't do much, as these functions don't have much in common

[jorg-vr]

@coderabbitai review

[bmesuere]

@coderabbitai review

[jorg-vr]

rebase and request review