[TohruSu]

AdaptiveQueue now supports reproducible experiments comparing scheduling policies on a single non-preemptive worker. Policies receive (waiting_jobs, current_time) with currently available jobs and the current integer simulation time; the episode continues until every job finishes. Baseline selection rules and seeded Random behavior are unchanged.

• Add immutable job specifications, per-job execution records, and an integer-time simulation engine.
• Implement FIFO, SPT, EDF, Priority, and uniformly seeded Random scheduling, with arrival/ID tie-breaking for deterministic policies.
• Add configurable stable, bursty, and mixed workloads (60 jobs by default), priority-weighted scheduling cost, deadline miss rate, and waiting/flow metrics.
• Provide an installable CLI, practical README, simulation reference, and CI checks on Python 3.10 and 3.12.

The clock advances to completions or the next arrival instead of processing ticks with no decisions. Workloads and Random policies use separate local RNGs. The waiting-time 95th percentile uses linear interpolation; empty episodes have zero metrics.

Validation on Python 3.12: 39 pytest tests passed; Ruff lint and formatting checks, strict mypy targeting Python 3.10, and git diff whitespace checks passed. Both README example commands ran successfully.