Improve presentation performance

[borkdude]

Improves presentation performance by

• avoiding postwalks for prepend-required-cljs and process-blobs
• caching the image processing in memory
• lowering the bounded count limit from 1.000.000 to 1.000

Closes #731.

[borkdude]

@mk Benchmark running 20x book.clj rendering against main vs this PR:

⏺ Now we see a real difference!
  ┌────────┬─────────────┬────────────────┐
  │ Metric │ Main branch │ cljs-deps-perf │
  ├────────┼─────────────┼────────────────┤
  │ Avg    │ 483.2ms     │ 433.4ms        │
  ├────────┼─────────────┼────────────────┤
  │ Median │ 474.8ms     │ 419.4ms        │
  ├────────┼─────────────┼────────────────┤
  │ Min    │ 414.4ms     │ 392.1ms        │
  ├────────┼─────────────┼────────────────┤
  │ Max    │ 683.9ms     │ 586.8ms        │
  └────────┴─────────────┴────────────────┘
  That's a ~10% improvement overall, and the presentation time is dramatically lower — look at the steady-state numbers:

  - Main: presentation ~150-160ms
  - Optimized: presentation ~70-85ms

  Presentation time is roughly halved. The overall improvement is diluted by parsing and eval time which are unchanged.

[borkdude]

@mk Ready to merge on my behalf

[jackrusher]

@borkdude Claude Code output? 😆

[borkdude]

@jackrusher yes, I told it to run my benchmark agains the current branch and a REPL that was opened on the main branch ;)