[Perf] Avoid repeated multimodal prompt update scans

[gty111]

Purpose

Avoid quadratic prompt-update planning when many multimodal items share the
same target.

For non-empty replacements, the previous implementation applies one item per
round. Each round still walks all unresolved items and rebuilds the match list,
resulting in roughly O(N²) item processing.

This change compiles updates with identical ordered (mode, target) choices
into FIFO queues. Matching only examines the next item in each queue and
consumes items in their original priority order. The common shared-target path
becomes O(prompt length + output length + item count).

Alternative targets, insertion and replacement modes, PromptIndex,
overlapping matches, and modality priority retain their existing behavior.

This is not a duplicate: no open PR addresses repeated shared-target update
rounds; #50716 optimizes individual scans but does not remove the per-item
rounds.

AI assistance was used.

Performance

Synthetic apply_token_matches benchmark:

• 10,000 image placeholders
• 1,000,000 input tokens
• 50 replacement tokens per image
• 1,490,000 output tokens

Implementation	Time
Previous implementation	85.97 s
This PR	0.086 s

This is approximately a 995x speedup. The expanded tokens and match results
are identical.

Test Plan

.venv/bin/python -m pytest tests/multimodal/test_processing.py -q
.venv/bin/python -m pytest tests/models/multimodal/processing/test_common.py \
  -k 'Qwen2-VL or llava-1.5 or SmolVLM2-2.2B' -q
pre-commit run ruff-check --files \
  vllm/multimodal/processing/processor.py tests/multimodal/test_processing.py
pre-commit run ruff-format --files \
  vllm/multimodal/processing/processor.py tests/multimodal/test_processing.py
pre-commit run mypy-3.12 --files \
  vllm/multimodal/processing/processor.py tests/multimodal/test_processing.py \
  --hook-stage manual

Test Result

• 95 processing tests passed.
• 9 model processing tests passed.
• 20,000 randomized differential cases matched the previous behavior.
• All staged pre-commit hooks passed.

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[DarkLight1337]

Thanks, LGTM now

[DarkLight1337]

/ci run

[github-actions]

✅ Triggered Buildkite CI #83315 for commit c390e7c284c8.