Rasterize layered map updates for faster composition

[SamuelFoo]

New feature implementation

Implemented feature

This PR follows up on SamuelFoo/next_gen_prototype#3 by rasterizing region updates when they arrive instead of retaining region geometry and rasterizing it again during every map composition.

The server stores contributions by source, map, update type, and grid cell.

Implementation description

This PR adds:

• transformation and rasterization of region updates
• per-source cell contributions with TTL and update ordering
• preservation of clear-before-obstacle and reset_source behavior
• removal of expired contributions without re-rasterizing regions
• invalidation of dynamic cells when static-map geometry changes
• a bounded FIFO queue for updates received before the static map
• tests for refresh, expiry, source ordering, geometry changes, and queue limits
• deterministic reset, rolling-overlap, and rolling-moving benchmarks

Benchmark results

• Baseline: 4d121ce (feat/replan-obstacle-demo)
• Candidate: 36c8099 (feat/rasterize-region-updates)
• Workload: 240 updates × 8 measured rounds, 2 warm-up rounds
• Input: 90 beams from 2 sources
• Checksums match

Scenario	Revision	Throughput (updates/s)	Peak RSS (MiB)	Checksum
reset	Baseline	581.6	6.75	124660162742684993
reset	Candidate	498.0	9.25	124660162742684993
rolling-overlap	Baseline	34.9	7.16	14420565338181443581
rolling-overlap	Candidate	648.9	9.24	14420565338181443581
rolling-moving	Baseline	34.7	7.16	8736190734789952998
rolling-moving	Candidate	559.6	11.32	8736190734789952998

Run the benchmark scenarios with:

ros2 run rmf_layered_map_server layered_map_benchmark \
  --label candidate --scenario reset

ros2 run rmf_layered_map_server layered_map_benchmark \
  --label candidate --scenario rolling-overlap

ros2 run rmf_layered_map_server layered_map_benchmark \
  --label candidate --scenario rolling-moving

Matching checksums confirm that two revisions produce the same composed maps before their performance results are compared.

GenAI Use

We follow OSRA's policy on GenAI tools

• I used a GenAI tool in this PR.
• I did not use GenAI

Generated-by: GPT-5

@arjo129 ready for review!