Add replay-based warm starts and time-varying capacity

[JaeseungYeom]

• add non-preemptive capacity schedules with scheduler, accounting, tracing, backfill, custom-scheduler, CLI, protobuf, gRPC, and Python integration
• implement replay-based warm starts using a temporary historical-event stage followed by an overhead-free normal simulation stage
• preserve historical warm-job departures while applying run-time policy only to post-boundary simulated jobs
• suppress historical seed jobs from output and statistics while retaining their resource and policy state
• enforce max_time as an inclusive absolute boundary in simulation, replay, progressive-loading, and warm-start execution
• default timestamp_format to epoch, auto-detect timestamp encoding per input, and reject mixed or invalid timestamp data
• validate simulation and replay runtime fields and improve trace parsing
• add heuristic capacity-period detection, comprehensive queue-pause and backfill analysis tools, simulator-compatible capacity output, and plotting
• retain the synthetic initialization-trace helper as a documented legacy workaround while recommending replay-based warm starts
• add capacity, warm-start, max-time, trace-tool, API, and gRPC test coverage
• update CMake, CI inventories, examples, and user/developer documentation