[ntuple][rfile] Add support for RFile in RNTupleWriter

[silverweed]

This Pull request:

allows RMiniFileWriter to handle RFile and exposes an experimental way to create an RNTupleWriter using RFile.
Currently the semantics are the same as the TDirectory overload: you pass in a reference to RFile and are responsible to keep the file alive until the RNTupleWriter is done.

Main question to answer: is this the semantics we want? It seems to me that it's the most straightforward API in C++, but it requires some extra work in python to avoid use-after-free cases like:

def CreateWriter():
  file = ROOT.RFile.Open(...)
  writer = ROOT.Experimental.RNTupleWriter_Append(..., file)
  return writer  # <- file may be GC'd from here on

This should be solveable by pythonizing Append to hold on to the file's proxy on the python side, but there might be additional complications I'm not considering. I'm open for discussion.

Checklist:

• tested changes locally
• updated the docs (if necessary)

[github-actions]

Test Results

    22 files      22 suites   3d 15h 11m 1s ⏱️
 3 770 tests  3 719 ✅ 0 💤 51 ❌
74 986 runs  74 934 ✅ 0 💤 52 ❌

For more details on these failures, see this check.

Results for commit 11610e1b.

♻️ This comment has been updated with latest results.

[hahnjo]

LGTM code-wise, I'll let Jakob approve the interface part (which I guess you already discussed offline)

[jblomer]

LGTM.

I think we should have a follow-up PR to rename the internal file structs in the mini file writer, e.g. RImplSimple, RImplTFile, RImplRFile... or something better. It just seems that currently the code overuses the word File.

[silverweed]

Added a python test (and a small pythonization of the RFile-based Append)