refactor: remove dead code and share test setup

[lukecotter]

📝 PR Overview

Removes code the extension no longer runs, and gives the test suites shared setup.
Nothing a user can reach changes.

This is the first of two planned reduction PRs. This one is deletions and test-helper
extraction only, so it cannot alter behaviour. A second PR will share duplicated
production code — the three database grid views, the find and column-view controllers —
and is deliberately kept separate because those diffs touch live paths.

Draft: in progress. More commits are coming on this branch. Each one is a single
theme and lands with pnpm lint and pnpm test green.

🛠️ Changes made

Landed so far:

• Drop the sprite timeline render path. The mesh renderers replaced it: FlameChart
  uses MeshRectangleRenderer, MeshMarkerRenderer and MeshAxisRenderer, and
  SearchOrchestrator uses MeshSearchStyleRenderer. No production file imported the
  sprite path. Deletes AxisRenderer (287), SearchStyleRenderer (209),
  TimelineMarkerRenderer (192), SpritePool (154), EventBatchRenderer (144) — whose
  only importers were each other — plus markers.test.ts (699) and batching.test.ts
  (564), which constructed only those classes.

Still to come on this branch:

• Drop TimeGridCalculator.ts, which has no importer.
• Drop dead exports, two dead shims and three dead CSS selectors.
• Drop the unused @rollup/plugin-alias devDependency.
• Map #vscode-elements in the jest config, removing 35 copies of the same
  jest.mock(…) line.
• Shared test helpers: a lit mount/settle pair (28 local copies today), log event
  builders (4 near-identical createEvent clones), apex-log fixtures (the same
  EXECUTION_STARTED line pasted across 22 files), tabulator and pixi doubles, and the
  lana asContext cast (67 occurrences).

🧩 Type of change (check all applicable)

• 🐛 Bug fix - something not working as expected
• ✨ New feature – adds new functionality
• ♻️ Refactor - internal changes with no user impact
• ⚡ Performance Improvement
• 📝 Documentation - README or documentation site changes
• 🔧 Chore - dev tooling, CI, config
• 💥 Breaking change

📷 Screenshots / gifs / video [optional]

No UI change. The timeline renders through the mesh path before and after.

🔗 Related Issues

None.

✅ Tests added?

• 👍 yes

sortMarkersByTimeAndSeverity had no coverage outside the deleted markers.test.ts, so
it gains a case in MarkerProcessor.test.ts. The later commits add shared helpers rather
than new cases.

Coverage this PR gives up. markers.test.ts also covered marker viewport culling,
end-time resolution and per-type colour. Those cases asserted sprite tint, alpha, x
and width — internals of the deleted class — so they could not move across as written.
MeshMarkerRenderer keeps visibleIndicators private and had no suite of its own before
this PR either, so the gap is pre-existing rather than new, but it is not closed here.
The fix worth making is extracting MeshMarkerRenderer.render()'s pure per-marker block
into MarkerProcessor, where those three behaviours become directly testable. Filed as
follow-up, not done here.

batching.test.ts needs no replacement: TemporalSegmentTree.test.ts and
RectangleCache.bucket.test.ts already cover the culling, size classification and
bucketing it exercised, against the live path.

📚 Docs updated?

• 🙅 not needed

No user-visible change, so no CHANGELOG entry. README and help site untouched.

Anything else we need to know? [optional]

Verified per commit with pnpm test, pnpm lint and pnpm build. The build is the
load-bearing check for the deletions — rollup has to resolve every import for the bundles
to be produced.

QuickPickWorkspace.ts has no caller and is kept on purpose: it holds the
multi-root workspace prompt we want to reinstate. It now carries a comment saying so, and
naming RetrieveLogFile.ts:68 and LogView.ts:213 as the two sites that take
workspaceFolders[0] today.

Three things found while reviewing, all left alone as out of scope:

• MetricStripRenderer.ts:471 re-sorts the marker array every frame
  ([...markers].sort) inside the 16.6ms budget, and exception markers are unbounded.
  Memoising on array identity is a three-line fix.
• SearchHighlightRenderer.ts:150 hardcodes const EVENT_HEIGHT = 15 where its twin
  uses TIMELINE_CONSTANTS.EVENT_HEIGHT, which that file already imports.
• The frame inset derivation (halfGap, gappedHeight) is repeated across five sites,
  and the mesh ones clamp with Math.max(0, …) where HighlightRenderer does not. The
  "must match" comments describe that coupling instead of enforcing it.

Out of scope by agreement: apex-log-parser/ is untouched, and the legacy timeline
behind lana.timeline.legacy is a separate piece of work.