(cherry picked from commit 152185c8a9de67a2b4f6913ee8deda6fef5a7448)

The Jupyter config falls back to the CMAKE_INSTALL_PREFIX/js folder to
find JSROOT if ROOTSYS is not set. This ensures installations in prefix
directories to find the installed JSROOT libraries.

Add jupyter_notebook_config.py to JUPYTER_CONFIG_DIR set in the ROOT
setup script, which allows now to use "jupyter notebook" with the same
capabilities than "root --notebook".

- Add build option `asan` to build with address sanitizer.
- Add a suppression file etc/lsan-root.supp with leak sanitizer suppressions.
- Add a directory core/sanitizer.
  - Add a micro library for configuring asan/lsan in ROOT. It is statically linked
    into all ROOT executables.
  - Create a target ROOT::ROOTStaticSanitizerConfig that exectuables can
    link against. This automatically pulls in the correct linker interface.
    This allows to create own sanitised executables that work with a
    sanitised ROOT.
  - Add a shared version of the config library for all cases where
    the address sanitizer run time is LD_PRELOADed into non-instrumented
    executables.
  - Add a README.md with more details.

Idea to implement only few methods for display of TGeoVolume 

add extra static paths to JSROOT scripts

use such static location of JSROOT scripts

release notes aboud offline usage

Wim says: adding 'shell=True' makes the command run on
/bin/sh and allows you to use shell features (such as expansion and
replacement of variables). But nothing in the command seems to need that
feature, and anyway it's a risky thing to rely on, or what am I missing?

Spawning an intermediate shell makes a mess of the environment when running
under conda on Mac in the non-build environment. Running the normal way as
per the patched line above, and all seems good.

cling now uses msync for the pointer check, not write.

The PCH configuration should match the runtime environment of ROOT so that
whenever ROOT starts up, the PCH should match the precise invocation options
of cling. In the past that was done by exporting plenty of build information
from the build system to the pch generation (done via rootcling). However,
this is not necessary anymore because rootcling and cling are much better
fine tuned towards detecting the build environment and setting the runtime
invocation options. Essentially, cling setup nowadays duplicates to some
extent this logic deeming it needless. Verification of this claim is quite
simple: ROOT will not start if the PCH build options mismatch.

This patch simplifies the pch generation logic by relying more on the cling
automatic detection of the build options.

See commit a07af0af

libPeac was removed in version 5.34.01

This is done by taking the check of fUniqueID and the assignment to fBuits
into a separate function and by including this function in the header
*but* marking it as "noinline" to prevent its inlining and preserve a
way to suppress the valgrind warning.

Introducing the macro R__NEVER_INLINE which expands to

   inline __attribute__((noinline))

The performance degradation is less than 10% for a case doing
only creation and destruction of TObjects.

etc was not appropriate place, one need to provide rules to install
JSROOT and OPENUI5 in <prefix> directory

Add possiblity to set collection in the table view. Fixes related to collection filtering and event loops.

Just for debugging, let stop ROOT and close browser window

plus several bug fixes

Also when toggling parent visisility, try to add/remove childs from
drawing

Many nodes in geometry may refer to same volume, therefore all nodes
should be checked when volume visibility flag is changed

When not full stack match is achieved, best shown with grey color