feat: add PHYSLITE ROOT file loader for ATLAS Open Data (#823)

[rx18-eng]

Title: feat: add PHYSLITE ROOT file loader for ATLAS Open Data

Body:

Closes #823

Adds a loader for ATLAS DAOD_PHYSLITE ROOT files, allowing users to browse
real ATLAS Open Data directly in Phoenix via the IO options dialog.

What it does

Reads calibrated physics objects from .root files using jsroot's TSelector
API and converts them to Phoenix event format. Supports: Jets, Tracks (InDet,
Muon Spectrometer, Combined, Extrapolated, GSF), Electrons, Muons, Photons,
MET, Vertices, and Calorimeter Clusters.

The loader checks branch existence before reading, so it works with both real
data and MC files automatically.

Tested with

• data15 minimum-bias (27 MB, 18,571 events) — MET, vertices, tracks load
  correctly
• MC ttbar simulation (613 MB, 30,000 events) — ~19 jets, ~115 tracks per
  event, electrons, muons, photons all display correctly
• No THREE.js NaN errors after guard fixes in phoenix-objects.ts

Files

• loaders/physlite-loader.ts — new loader (reads CollectionTree, computes
  energy from pt/eta/phi/m)
• loaders/objects/phoenix-objects.ts — NaN guards for track extrapolation
  and calorimeter geometry
• index.ts — export
• io-options-dialog.component.ts — PHYSLITE import option
• event-data-import.ts — PHYSLITE enum value
• atlas.component.ts — enabled for ATLAS section

Recording.2026-03-16.225316.mp4

[rx18-eng]

Hey @EdwardMoyse @sponce , please let me know your thoughts on this

[deveshbervar]

Hi @EdwardMoyse,

I've been independently investigating issue #823 in parallel
and tested against the same PHYSLITE file from CERN Open Data
(DAOD_PHYSLITE.37001626._000001.pool.root.1, data15, 18,571 events).

I can confirm:

• jsroot opens PHYSLITE successfully (no compression issues unlike Investigate opening ATLAS AOD files directly via jsroot #44)
• treeProcess + TSelector is the correct API approach
• 397 branches readable across 18,571 events
• Jets, Tracks, MET, Vertices, CaloClusters branch mappings
  are all correct.

One issue I found while verifying branch names against
the actual file:

The loader expects AnalysisElectronsAuxDyn.pt and
AnalysisMuonsAuxDyn.pt but these branches do not exist
in the data15 file. What actually exists is:

• AnalysisElectronsAux. (no AuxDyn sub-branches)
• AnalysisMuonsAux. (no AuxDyn sub-branches)

This means Electrons and Muons will be silently skipped
for this file due to the branch existence check. The PR
description mentions testing with MC ttbar — it's possible
AuxDyn branches exist in MC but not in data15 real data files.

This might also be specific to the data15 file I tested
with — would be great to confirm against the MC file
used in testing.

Happy to help investigate further or test additional files.

[rx18-eng]

Hey, thanks for testing! This is actually already handled , the loader checks branch existence before adding anything to the TSelector, so collections like Electrons/Muons get skipped gracefully when their AuxDyn branches don't exist in the file.

You're right that the data15 file stores them as serialized blobs (AnalysisElectronsAux.) instead of flat AuxDyn branches. I ran into the same thing during development. The MC ttbar file does have the AuxDyn branches and
electrons/muons load fine there.

So it's working as intended , the loader adapts to whatever branches are available in the file.