Add Rust and Python video tracking E2E demo

[rsasaki0109]

Summary

• add a deterministic Rust/Python frame-sequence demo that reloads 12 PGM frames, verifies dense optical flow for two opposite motions, and preserves native tracker IDs
• expose the existing safe Rust MultiObjectTracker as a stateful Python class with lossless typed track tuples and matching stubs/tests
• generate and commit the 12-frame GIF, plus README, ROADMAP, CI smoke coverage, and a dated reproducibility receipt

Validation

• cargo test -p spatialrust-vision --no-default-features --features video (22 passed)
• cargo run -p spatialrust --no-default-features --features image-io-standard,vision-video --example video_tracking_e2e
• cargo check --manifest-path crates/spatialrust-py/Cargo.toml
• cargo clippy --manifest-path crates/spatialrust-py/Cargo.toml -- -A clippy::missing-const-for-thread-local -A clippy::too-many-arguments -D warnings
• maturin develop --release --manifest-path crates/spatialrust-py/Cargo.toml
• .venv/Scripts/python.exe -m pytest crates/spatialrust-py/tests -q
• .venv/Scripts/python.exe -m mypy.stubtest spatialrust --ignore-missing-stub
• .venv/Scripts/python.exe crates/spatialrust-py/examples/video_tracking_e2e.py

Reproducibility

Rust and Python independently generated 12 byte-identical 96x72 Gray8 PGM frames. All 11 frame pairs produced exact object-center flow (2,1) and (-2,-1), detections remained at two objects, and track IDs remained 1,2. The committed GIF is 384x288 with 12 frames at 140 ms/frame.