[tree] prevent nullptr access in ttreereader

[ferdymercury]

This Pull request:

Changes or fixes:

nullptr access when branch is not found

Fixes https://its.cern.ch/jira/browse/ROOT-8842 by @jpivarski

Checklist:

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

[github-actions]

Test Results

    21 files      21 suites   3d 14h 10m 46s ⏱️
 3 787 tests  3 787 ✅ 0 💤 0 ❌
77 580 runs  77 580 ✅ 0 💤 0 ❌

Results for commit 4f04fe5.

♻️ This comment has been updated with latest results.

[dpiparo]

Thanks for this improvement. I propose the following way forward, if you agree:

1. A small unit test is added for this fix
2. We wait for @pcanal to approve (with a test, I would be ready to approve and to pick up the rest of the fixing, but I prefer to have his opinion on this)
3. We promote the old JIRA item as a new GH issue specific to the residual issue

Does it make sense?

[ferdymercury]

The issue seems finally fixed now (tried locally).

Side note, the equivalent function in TTreeReader(NotFast) has:

if (fProxiesSet) {
      Error("RegisterValueReader",
            "Error registering reader for %s: TTreeReaderValue/Array objects must be created before the call to Next() "
            "/ SetEntry() / SetLocalEntry(), or after TTreeReader::Restart()!",
            reader->GetBranchName());
      return false;
   }

[dpiparo]

Let's see how the builds go after the rebase to the current main branch. If the builds are green, it's just a matter of removing the commented code (I can do that, no worries) and then for me the code is ready to go.