(cherry picked from commit 2857e502)

XWIKI-22917: Solr indexing takes very long when there are many objects with different values (#3923)

* Use a HashSet to compute unique values instead of a linear search.
* Rename LengthSolrInputDocument to XWikiSolrInputDocument.
* Use XWikiSolrInputDocument in more places to use the new method.
* Add a unit test for XWikiSolrInputDocument.

(cherry picked from commit 39e2c7ff)

XWIKI-22583: Allow users with edit rights on a page and all its children to use the Pinned Pages feature

  * Fix flickering test

(cherry picked from commit 0b95939a)

  * Fix flickering test

(cherry picked from commit 93fafb7d)

XWIKI-22926: Make it possible to configure the batch size in the Solr index synchronization on startup (#3928)

* Add a new configuration property for configuring the batch size of the
  initial Solr indexing synchronization.
* Update and add tests.

(cherry picked from commit db2814b6)

(cherry picked from commit fa46bfe8)

* Don't do a linear search for finding the object with the correct number.

(cherry picked from commit 1bf17199)

* Fix trim being applied to the wrong value.

(cherry picked from commit 56158106)

XWIKI-22583: Allow users with edit rights on a page and all its children to use the Pinned Pages feature (#3900)

*  Provide a REST API to manipulate pinned pages with Edit right only
* Provide an index script service and an index.tree script service to manipulate pinned pages
* Provide tabs in children viewer with a UIXP
* Use the UIXP to inject capability to configure pinned pages reusing the navigation panel administration UI
* Provide new integration test

(cherry picked from commit aaf345bb)

  * Revert changes related to aria-hidden and tabsindex: we want user to
    be still able to use keyboard to pin/unpin pages

(cherry picked from commit 3ca4f633)

- Replace & with && in if
- Remove superfluous generics definitions
- Use new API for maps manipulation

(cherry picked from commit 805e91db)

Co-authored-by: Manuel Leduc <manuel.leduc@xwiki.com>

* Ensure that the specified fields are actually valid field names
* Don't allow password and email fields
* Add tests

(cherry picked from commit f2ca8649)

* Solved the unlabelled button issue.
* Solved the anauthorized role final node issue.

(cherry picked from commit 1f8019f2)

  * Fix javadoc

(cherry picked from commit 5bd54373)

* bulletproof the argument displayer to catch any exception
* use a more reliable way to identify object types

(cherry picked from commit 863e5f54)

  * Ensure to always send an empty value so that it's always an array
    sent to the server

(cherry picked from commit 66ca574f)

(cherry picked from commit 542b1243)

(cherry picked from commit fc7b7943)

* Merging consecutive PDF link annotations with the same target doesn't make sense all the time (this PO is used to test the old PDF export as well).

(cherry picked from commit 7f154d4c)

* improve internal API

(cherry picked from commit dcc9ac20)

* fix test

(cherry picked from commit 6984a161)

* fix XWikiDocument#getXClass() causing the document to become dirty on Oracle

(cherry picked from commit 0549a826)

* Treat stylesheets loaded from a different domain as missing because paged.js fails to load them anyway without recoving (blocking the PDF export). We'll have to rollback this changed once https://github.com/pagedjs/pagedjs/issues/267 is fixed.

(cherry picked from commit 203d57c9)

(cherry picked from commit e0d43bd5)

(cherry picked from commit f0302555)

[stable-16.4.x] XWIKI-22782: Only save modified xobjects (#3899)

(cherry picked from commit 0587cb0c)

* Workaround for a paged.js bug

(cherry picked from commit d0a24681)

* Mark group objects as clean after loading them.

(cherry picked from commit c83e77a7)

* Add a new MacroParameterRequiredRightsAnalyzer role.
* Add implementations for wiki content and macro content source references.

(cherry picked from commit 4dd774ce)

  * Ignore returning the value of ComputedField properties in REST API,
    since by definition those field value is computed.

(cherry picked from commit 79032f58)

(cherry picked from commit cc8fa85e)

(cherry picked from commit 4992bbba)

(cherry picked from commit 19fa9530)

(cherry picked from commit a403c9f9)

XWIKI-22791: Costly and unconditional tag cloud computation done for all live table requests using the default source

(cherry picked from commit 85bb58df)

XWIKI-22836: Author is not properly set when refactoring links

  * Fix integration test in that branch

* Configure the "string" field type with docValues="true" as in the
  default configuration. In tests, this eliminates all fields in the
  field cache apart from the caches for title_sort and
  author_display_sort, which seem small in comparison.

(cherry picked from commit 062dabeb)

(cherry picked from commit 84e1efe4)