[visualizer] feat: add memory timeline HTML visualizer

[tifning]

What does this PR do?

Adds a Plotly-based interactive memory timeline visualizer (memory_html) that renders
NPU memory allocation events as an interactive Gantt chart. Key features:

• Dual-chart layout: Chart1 shows global memory usage trend (line + area fill); Chart2
  shows per-operator Gantt bars, color-coded by operator name with bidirectional zoom sync.
• Time-window segmentation: Large datasets (100k+ events) are split into at most 20
  time-window segments, each generating a standalone HTML+JS file pair with Prev/Next
  navigation. Chart1 data (global timeline) is shared across all segments.
• Segment smart hint: When the user zooms Chart1 into a different time range, a "Best:
  Seg N →" hint appears with a clickable link to the matching segment file.
• Overlap detection: When multiple bars of the same operator stack at the same time
  point, the hover tooltip and detail panel both list all overlapping events with size and
  timing info. Click any row in the detail panel to navigate between overlapping events.
• Call stack display: call_stack field is stored per event and shown in the detail
  panel on bar click (empty call stacks display as (empty) gracefully).
• Absolute time display: Chart1 x-axis tick labels and hover tooltips show absolute
  timestamps, while internal data uses relative times for compact JSON output.

Checklist Before Starting

• Search for similar PRs. Paste at least one query link here: ...
• Format the PR title as [{modules}] {type}: {description} (This will be checked by the CI)
  • {modules} include pipeline, parser, visualizer, data, deployment, perf, algo, env, doc, cfg, ci, misc
  • If this PR involves multiple modules, separate them with , like [mstx, ci]
  • {type} is in feat, fix, refactor, chore, test
  • If this PR breaks any API (CLI arguments, config, function signature, etc.), add [BREAKING] to the beginning of the title.
  • Example: [BREAKING][mstx, torch_profile] feat: support timeline parsing

Test

python -m rl_insight.main input.path=xx timeline.parser.type=memory timeline.visualizer.type=memory_html output.path=xx

For changes that can not be tested by CI (e.g., algorithm implementation, new model support), validate by experiment(s) and show results like training curve plots, evaluation results, etc.

API and Usage Example

Demonstrate how the API changes if any, and provide usage example(s) if possible.

# Add code snippet or script demonstrating how to use this

Design & Code Changes

New files:

• rl_insight/visualizer/memory_visualizer.py — MemoryVisualizer class registered as
  memory_html cluster visualizer. Handles data preprocessing, segment splitting,
  JSON generation, and HTML template injection.
• rl_insight/visualizer/memory_template.html — Plotly-based HTML template (~3 KB)
  with all rendering logic in vanilla JS. Data lives in a separate detail_data.js file.

Modified files:

• rl_insight/visualizer/__init__.py — Added from .memory_visualizer import MemoryVisualizer
  to register the memory_html visualizer.

Key design decisions:

• Data arrays are stored in external .js files (not inlined in HTML), keeping the HTML
  template lightweight and cacheable.
• call_stack strings use a pool+index pattern (CS_POOL / CS_IDX) for deduplication,
  minimizing JSON output size.
• Operator bars are grouped into a single Plotly trace per operator using x:[] and
  base:[] arrays, reducing trace count from O(events) to O(unique_operators).
• Segment overlap uses start + duration > seg_start AND start < seg_end so events
  spanning segment boundaries appear in both files.

Checklist Before Submitting

Important

Please check all the following items before requesting a review, otherwise the reviewer might deprioritize this PR for review.

• Read the Contribute Guide.
• Apply pre-commit checks: pre-commit install && pre-commit run --all-files --show-diff-on-failure --color=always
• Add / Update the documentation.
• Add unit or end-to-end test(s) to the CI workflow to cover all the code. If not feasible, explain why: ...

[gemini-code-assist]

Code Review

This pull request introduces a new MemoryVisualizer and an accompanying HTML template to generate interactive memory allocation timelines and operator Gantt charts. The feedback focuses on critical performance optimizations and a potential bug fix: replacing an $O(N \times M)$ nested loop in _build_chart1_data with a sweep-line (two-pointer) approach, vectorizing or optimizing DataFrame iterations to avoid slow iterrows() calls, and handling relative paths correctly in directory creation to prevent a FileNotFoundError.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[tardis-key]

As a complete feature, we also need the data (/data), the datachecr(/rl_insight/data/data_checker.py)the data documentation (/docs/data), the examples (/examples), e2e test (tests/special_e2e).