feat: add in-browser histogram panel using jsroot (#840)

[rx18-eng]

Summary

• added histogram panel to the toolbar — shows invariant mass distributions live using jsroot
• students tag particles, compute mass, histogram fills in real-time (the Z peak at ~91 GeV emerges as they work)
• event bus on EventDisplay (on/emit) connects the masterclass panel to the histogram

What changed since first push

• dark theme fix — gStyle is now saved/restored around jsroot draw calls so it doesn't leak into geometry rendering
• no more ngDoCheck — replaced with proper @input setter, was firing every change detection cycle
• removed ViewEncapsulation.None — styles no longer leak globally, uses scoped ::ng-deep
• removed Higgs hint line — Z-path masterclass data won't have Higgs events, a line at 125 with nothing near it confuses
  students. added separate ATLAS_JPSI_HISTOGRAM and ATLAS_WIDE_HISTOGRAM configs for other exercises
• configurable event name — config.eventName defaults to 'result-recorded', LHCb can use 'd0-lifetime-recorded' or whatever they
  want
• HistogramConfig threaded via @input — ui-menu → histogram-panel → overlay, same pattern as masterclass config. experiments pass
  their own config
• localStorage persistence — histogram data survives page refresh/crash. raw values stored and restored on load. critical for
  3-hour masterclass sessions
• reset confirmation — asks "Clear all N entries?" before wiping, prevents accidental data loss
• debounced redraw — 50ms coalesce so bulk-loading 100+ events doesn't cause lag
• hint labels show units — Z (91.2 GeV) instead of just Z
• TSV export — tab-separated instead of comma, avoids European locale decimal issues. filename derived from config title
• brighter fill color — ROOT color 857, more visible on dark background
• narrowed default range — 20-120 GeV instead of 0-150, Z peak is front and center

Test

Simulated a realistic Z-path masterclass session in the browser console ->175 events in 3 phases:

• Phase A (40 events, 200ms each): careful student pair, mostly correct Z→μμ and Z→ee with realistic detector resolution, a few wrong combinations and off-shell Z*
• Phase B (35 events, 250ms each): less experienced pair, more mistakes — same-sign lepton pairs, jets misidentified as leptons,boundary/overflow values
• Phase C (100 events, instant): instructor bulk-loads pre-analyzed results to test debounce
  Uses Breit-Wigner + Gaussian smearing for realistic Z line shape. Includes wrong combos, out-of-range values, and edge cases.

video link

Closes #840

[rx18-eng]

hey @sponce @EdwardMoyse , this adds the histogram panel from #840. @cohm mentioned he wants to test this with students in Stockholm on Monday so it would be great if you could take a quick look before then. happy to make changes if anything needs fixing!

[EdwardMoyse]

Sorry for taking so long to review this ... but it seems like it is not clean and pulls commits in from other PRs e.g. #831 ?

[EdwardMoyse]

Thanks for the update @rx18-eng - I confess I'm still a bit confused about what is happening in your video (and unfortunately, as noted elsewhere I'm not really an expert in the masterclasses, but I will follow up). But how is the histogram being filled? You would not get a z peak from one event - are you feeding in a bunch of tracks from different events?

[EdwardMoyse]

@rx18-eng - just a quick ping about my question above... as I said, I'm a bit confused about what is going on in the demo.

[rx18-eng]

@EdwardMoyse I have loaded 175 events in 3 phases and the z peak is a result of that ! I f you say I can also share with you the sample code I used :) . Also great to see you after a long time :)

[rx18-eng]

Test

Simulated a realistic Z-path masterclass session in the browser console ->175 events in 3 phases:

• Phase A (40 events, 200ms each): careful student pair, mostly correct Z→μμ and Z→ee with realistic detector resolution, a few wrong combinations and off-shell Z*
• Phase B (35 events, 250ms each): less experienced pair, more mistakes — same-sign lepton pairs, jets misidentified as leptons,boundary/overflow values
• Phase C (100 events, instant): instructor bulk-loads pre-analyzed results to test debounce
  Uses Breit-Wigner + Gaussian smearing for realistic Z line shape. Includes wrong combos, out-of-range values, and edge cases.

This is the exact scenario I demonstarted in the video !