and test the feature.

- remove note on TIterationHelper renaming (it's internal)
- small clarification about support of `rdfentry_`, `rdfslot_`
- add mention that parallel execution is now saner

Provide methods to check if sending via websocket is possible
Fix - correctly handle negative parameter values in TF1/TF2
Fix - clear range flag when drawings are cleared
Fix - do not forget reset pads cache when changing TWebCanvas

Use the mathcore macro defined in Math/Error.h for error messages. This fixes a compilation problem observed on mac 10.14

Also increase arena size to speed up heavy allocations of RooArgSets.

this commits removes a substantial amount of code which made the
execution of core operations reentrant. Reentrancy is not a requirement
any more since the work spawned via TThreadExecutors is isolated, which
means that a task that submits subtasks, e.g. via a parallel_for, and
is idle waiting those subtasks to finish cannot pull in a work item
which was spawned by any of its parents.
Concretely, it means that while processing a cluster of entries in a
slot, it cannot happen that the processing is interrupted to start
processing a new cluster.

providing a data frame and a HEP framework example.

We enclose the parallel_for and parallel_reduce invocation in a
task_arena::isolate because we want to prevent work stealing.

re-expressing those in terms of a single internal template helper, therewith avoiding
duplication of code.

This work-around silents the error C2668: 'ROOT::TThreadExecutor::Map': ambiguous call to overloaded function, due to a MS compiler bug (as reported here: https://developercommunity.visualstudio.com/content/problem/225022/wrong-error-c2668-ambiguous-call-to-overloaded-fun.html)

- fix LZ4 error for large buffers
- add support of TProfile2Poly
- repair online context menu for TAxis
- consistent handling of tooltips on canvas level

Support gStyle and multiple palette transport from TWebCanvas
Better handle gStyle attributes for title drawing

We need changes because of a new format for compression settings.

The python and c++ versions of the roofit tutorials compile the same file,
which sometimes breaks one of the two when they run in parallel.