Openuniverse tutorial updates ahead of Sep 15 webinar

[jkrick]

Updates the OpenUniverse2024 tutorials to the full data release and reconciles how they find data.
Closes #309.

What changed

• All five notebooks use the full release, not the preview subset. Preview references are gone —
  a reader has no reason to know an earlier subset existed.
• Both telescopes are reached the same way. The firefly notebook built Roman paths from a hardcoded
  24-value block grid while asking SIA for Rubin. Both now go through SIA, removing the grid constants,
  block_wcs(), find_block(), get_roman_coadd_fpath(), and the grid-vs-header verification.
• Every static figure is rotated to north up. Neither survey delivers north-up images (this Roman
  exposure sits ~170° from north, the Rubin visit ~160°), so the four figures in §2/§3/§4/§6 were
  previously all at different orientations. Each is now resampled onto a north-up grid keeping its own
  pixel scale, so Roman's finer sampling stays visible. Whole exposures resample at coarser sampling for
  display. Rotation leaves empty corners, so stretch_color() is nan-aware. This is what reproject is
  for now; the old reproject_to_TAN() is gone.
• The supernova notebook replaces a hardcoded instances CSV with an SIA positional query, and gains
  playback controls and visit-coverage guidance.
• The Roman coadds notebook is deleted. See below.
• Reorganized landing page and toc: a nested OpenUniverse2024 group listed first, Quickstart at the
  top, prefix-free titles, simulated.md split into two galleries.
• Helper functions folded into hidden cells across all five notebooks, per Hide cells both in jupyterlab and myst rendering #335, and now carry numpydoc
  docstrings. Three cells were split first, since a cell folds as a unit.

Why no coadds

Neither survey's coadds are reachable through SIA for the full release.

• Roman WAS coadds exist in S3 but are not registered. IRSA is indexing the coadds, but
  not on a timeline that affects this PR.
• Rubin coadds were never generated. The preview did have one, which the original notebook used.

The notebooks use single exposures throughout and don't mention coadds, so nothing needs rewriting when the
Roman coadds are registered. Consequence for visualization notebook: every full-release Rubin product is a single visit, so the Rubin panel is genuinely shallow

CI cost

The firefly notebook was measured end-to-end, not estimated:

	data transferred	wall clock
before (coadd version)	4.9 GB	222 s
after	524 MB	229 s

Firefly: known issues, left for a separate PR

Firefly behaves inconsistently here. On some runs §5 and §7 work exactly as written — the Rubin image
displays and the region overlays appear — and on others the image does not display at all, on the same code
and inputs. I can't figure this out. This PR does not try to fix it. @jaladh-singhal please review sections 5 & 7 especially carefully, or open a new PR for any necessary changes.

firefly_client 3.4.0 also deprecated two things the notebook uses, so §5 and §7 print:

WARNING: show_fits() is deprecated. Use show_fits_image() instead.
WARNING: url is deprecated, use file_input parameter instead

Notes for reviewers

• I tried to finagle so that the PR will show diffs and not new files for the notebooks which were edited, not sure I succeeded.
• Also updated the top.yaml, but I can't figure out how to see what this looks like until I push.....so here goes.