This corresponds to the fix from Sergei Linev (1b565f24), which will
be integrated in cppyy-backend when the ROOT of Cppyy is updated to
a patch release of 6.16.

In the meantime, this patch reminds of the necessary fix.

In its last update (CPyCppyy 1.6.4), Cppyy automatically adds to
a derived class the overloads of a base class that come from
using declarations.

Therefore, that makes the pythonisation for plotOn redundant,
since everything will be handled at the bindings level.

If doing direct conversion of map -> json, it is not possible 
to specify "JSON_object" tag in class comments. Therefore now one can
use compact parameter 5 for that:

std::map<string,int> obj;
obj["name"]=7;
auto json = TBufferJSON::ToJSON(&obj,5);

Let stream classes std::map<int,double> directly

This instruct TBufferJSON to produce JSON object instead of array for
std::map classes

There are several non-trivial data members, therefore create only when
really needed. Especially such members are not needed during writing of
JSON. Should slightly improve performance

Members mapped with JSONmap comment, are stored as normal objects
and can be read as normal objects

If data member marked with "JSONmap" comment, such std::map class
will be converted in plain JSON object. Of course, first type in the
std::map MUST be of string (or TString or const char*). 
Such representation is much more compact than default one

This patch enables loading of dictionaries based on modules. As such, we have
a module which contains test/Event.h. ROOT now preloads it because it is in
the same folder with stressTMVA. Essentially it injects test Event class
on the global scope and using namespace TMVA enables the 'other event'.

I am not particularly happy with this workaround but disabling loading of
the module with the test event will harm more tests.

This patch actually loads the modulemap as previously it wrongly detected
it was already loaded.

Problem is visible only on CERN VM infrastructure and currently is under investigation, for 32b Docker container tests are passing.

Base classes for attributes defined in gpad, but also some classes like
RPad uses line or fill attributes. Later attributes can be extracted
into separate library.

This information is used by roottest-root-dataframe-
test_progressiveCSV to check the correct behaviour of the
progressive reading of the CSV data source.

this is achieved adding a special case which checks if the directory
containing the tree is a file or a proper directory.