chore: add startup-time plans

[cdc-as81]

Summary

This PR adds startup-time plans: one-shot lifecycle work that runs at the
effective simulation start time—either the configured start time or 0.0—
before regular plans.

• Add Context::add_startup_plan and
  Context::add_startup_plan_with_phase.
• Expose startup-plan scheduling through ContextBase and PluginContext.
• Expand the event-loop state machine to model startup, regular execution,
  shutdown, and abort/resume transitions explicitly.
• Add a dedicated startup queue with shared plan IDs, cancellation, and
  profiling support.
• Update the Context API documentation to describe the full simulation
  lifecycle.

Behavior

Startup-time plans:

• Run once at the effective simulation start time before any regular plan.
• Follow ExecutionPhase ordering, then insertion order within each phase.
• Yield to queued callbacks, which retain priority over plan selection.
• May schedule additional startup-time plans while startup is still active.
• Can be canceled through their PlanId.
• Resume on a later execute call when startup is interrupted by abort.
• Finish draining before shutdown proceeds when shutdown is requested during
  startup.
• Panic if added after startup has completed.

[RobertJacobsonCDC]

The issue described how to expand the event loop state machine in a way that expands it to startup time. This implementation uses a separate state tracking mechanism that obfuscates which states and transitions are possible. If there is a reason, I need an explanation.

Also, don't let the LLM factor out single lines of code into useless helpers. Introducing DistinguishedQueue, add_distinguished_plan, and pop_next_distinguished is pretty egregious.

[github-actions]

Benchmark Results

Hyperfine

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
large_sir::baseline	3.2 ± 0.0	3.2	3.3	1.00
large_sir::baseline_households	3.3 ± 0.1	3.2	3.5	1.01 ± 0.02
large_sir::entities	7.8 ± 0.2	7.6	8.9	2.41 ± 0.07
large_sir::households	7.5 ± 0.2	7.3	8.2	2.30 ± 0.05

Criterion

Regressions (slower)

Group	Bench	Param	Change	CI Lower	CI Upper
sample_entity	sample_entity_single_property_unindexed	10000	7.523%	7.115%	8.059%
large_dataset	bench_match_entity		5.139%	4.604%	5.863%
examples	example-basic-infection		2.579%	1.867%	3.265%
counts	reindex_after_adding_more_entities		1.315%	1.065%	1.591%

Improvements (faster)

Group	Bench	Param	Change	CI Lower	CI Upper
indexing	with_query_results_indexed_multi-property_entities		-13.466%	-14.264%	-12.629%
large_dataset	bench_filter_unindexed_entity		-10.806%	-13.938%	-7.714%
sample_entity	sample_entity_single_property_unindexed	1000	-10.004%	-12.243%	-7.497%
large_dataset	bench_query_population_multi_unindexed_entities		-9.883%	-11.580%	-8.084%
large_dataset	bench_query_population_indexed_property_entities		-7.938%	-8.656%	-7.279%
sampling	sampling_single_unindexed_entities		-5.111%	-5.397%	-4.808%
indexing	query_people_count_indexed_multi-property_entities		-4.826%	-6.306%	-3.086%
sample_entity	sample_entity_multi_property_indexed	100000	-4.627%	-5.222%	-4.049%
indexing	query_people_count_multiple_individually_indexed_properties_enti		-4.444%	-5.400%	-3.681%
large_dataset	bench_filter_indexed_entity		-4.359%	-6.291%	-2.396%
sample_entity	sample_entity_multi_property_indexed	1000	-4.323%	-4.750%	-3.926%
sample_entity	sample_entity_single_property_indexed	1000	-4.318%	-5.094%	-3.568%
sample_entity	sample_entity_multi_property_indexed	10000	-3.895%	-4.392%	-3.339%
large_dataset	bench_query_population_multi_indexed_entities		-3.863%	-5.614%	-2.261%
sample_entity	sample_entity_single_property_indexed	10000	-3.621%	-4.127%	-2.977%
sampling	sampling_multiple_unindexed_entities		-3.149%	-3.494%	-2.789%
sampling	count_and_sampling_single_known_length_entities		-3.108%	-3.567%	-2.722%
sampling	count_and_sampling_single_unindexed_concrete_plus_derived_entiti		-2.823%	-3.155%	-2.553%
sample_entity	sample_entity_whole_population	100000	-2.214%	-2.788%	-1.602%
sample_entity	sample_entity_single_property_indexed	100000	-2.198%	-2.712%	-1.621%
indexing	query_people_indexed_multi-property_entities		-2.010%	-2.333%	-1.687%
sampling	sampling_single_known_length_entities		-1.970%	-2.443%	-1.521%
sampling	sampling_multiple_l_reservoir_entities		-1.821%	-1.992%	-1.662%
indexing	query_people_multiple_individually_indexed_properties_entities		-1.720%	-2.371%	-1.166%
algorithm_benches	algorithm_sampling_multiple_l_reservoir		-1.639%	-2.119%	-1.188%

Unchanged / inconclusive (CI crosses 0%)

Group	Bench	Param	Change	CI Lower	CI Upper
indexing	with_query_results_single_indexed_property_entities		2.136%	0.214%	4.005%
sample_entity	sample_entity_whole_population	1000	-1.359%	-2.034%	-0.736%
sample_entity	sample_entity_whole_population	10000	-1.299%	-1.850%	-0.703%
counts	single_property_unindexed_entities		1.237%	-0.212%	3.574%
counts	multi_property_indexed_entities		1.085%	0.132%	2.209%
sampling	sampling_single_l_reservoir_entities		1.064%	0.647%	1.382%
sample_entity	sample_entity_single_property_unindexed	100000	-0.836%	-1.176%	-0.494%
algorithm_benches	algorithm_sampling_multiple_known_length		0.779%	0.249%	1.292%
counts	concrete_plus_derived_unindexed_entities		-0.750%	-1.505%	-0.121%
large_dataset	bench_query_population_property_entities		-0.749%	-1.444%	-0.145%
large_dataset	bench_query_population_derived_property_entities		0.428%	-0.281%	1.105%
indexing	query_people_count_single_indexed_property_entities		0.375%	-0.472%	1.267%
counts	single_property_indexed_entities		-0.373%	-1.115%	0.178%
sampling	sampling_multiple_known_length_entities		0.304%	-0.435%	0.818%
indexing	with_query_results_multiple_individually_indexed_properties_enti		-0.300%	-0.595%	-0.014%
indexing	query_people_single_indexed_property_entities		0.293%	-0.259%	0.738%
algorithm_benches	algorithm_sampling_single_l_reservoir		-0.289%	-0.766%	0.188%
examples	example-births-deaths		0.280%	-0.050%	0.622%
sampling	sampling_single_unindexed_concrete_plus_derived_entities		-0.275%	-0.450%	-0.108%
algorithm_benches	algorithm_sampling_single_known_length		-0.202%	-0.903%	0.537%
counts	index_after_adding_entities		-0.183%	-0.414%	0.044%
counts	multi_property_unindexed_entities		-0.122%	-0.980%	0.773%
algorithm_benches	algorithm_sampling_single_rand_reservoir		-0.122%	-0.359%	0.063%

Not Compared (no baseline yet)

Group	Bench	Reason
(none)

[cdc-as81]

The issue described how to expand the event loop state machine in a way that expands it to startup time. This implementation uses a separate state tracking mechanism that obfuscates which states and transitions are possible. If there is a reason, I need an explanation.

Also, don't let the LLM factor out single lines of code into useless helpers. Introducing DistinguishedQueue, add_distinguished_plan, and pop_next_distinguished is pretty egregious.

The separate startup-progress field is gone, and the lifecycle is now represented by a single explicit EventLoopStatus state machine. I also removed DistinguishedQueue, add_distinguished_plan, and pop_next_distinguished, and inlined the queue-specific logic. I added tests for abort and shutdown during startup, preserving pending shutdown across an abort, and single-step state transitions.