fix: Improve hygiene of trait impl macros

[RobertJacobsonCDC]

This PR improves the hygiene of macros that call trait methods on concrete types. Since we are touching this macro code anyway, we also normalize a particular syntax we use in macro definitions (described below), which is merely cosmetic and has no functional effect.

Use fully qualified trait method invocation in macro-generated trait impls when name collisions are possible

Suppose we have a macro that implements a trait for a concrete type identified by a macro parameter like $identifier, and that same macro also generates code that calls methods from that trait on that type. If those generated calls use ordinary syntax like $identifier::id() or $identifier::is_derived(), they can collide with inherent methods already defined on the concrete type. For example, if the macro implements Property<Person> for Age, and Age already has its own inherent fn id() -> usize, then generated code that says Age::id() may call the inherent method instead of the trait method Property<Person>::id(). And if Age::id() has a different signature from Property<Person>::id(), you get a compiler error.

We fix this by changing such calls to fully qualified trait method syntax, so the generated code explicitly names both the trait and the receiver type. Instead of generating $identifier::id() or $identifier::is_derived(), we generate calls like <$identifier as $crate::entity::property::Property<$entity>>::id() and <$identifier as $crate::entity::property::Property<$entity>>::is_derived(). This removes the ambiguity by telling Rust exactly which trait method to call, even if the concrete type also defines an inherent method with the same name.

This is motivated by the real example of trying to implement Property<Person> for FIPSCode, because the concrete type has a method FIPSCode::id that collides with Property::id.

Use consistent syntax for Self in fully qualified trait method invocation

This PR also normalizes fully qualified trait method calls so that, whenever the receiver is the current impl type, we write Self instead of repeating the concrete type name. For example, in src/macros/property_impl.rs we use <Self as $crate::entity::property::Property<$entity>>::id() rather than <$property as $crate::entity::property::Property<$entity>>::id().

The justification for this is:

• Consistency.
• Better with generics. Self avoids having to restate type parameters, lifetimes, and wrappers in the qualified path. That makes the macro simpler and less fragile.
• Better for refactors. If the concrete type spelling changes, Self keeps the call correct without needing to update repeated type syntax.
• Communicates intent more clearly. <Self as Trait>::... reads as "call the trait method from this impl."

This isn't a very important change, but it is also pretty low risk. Syntactic housekeeping since we are touching the code anyway.

[github-actions]

Benchmark Results

Hyperfine

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
large_sir::baseline	2.8 ± 0.0	2.7	3.0	1.00
large_sir::entities	12.4 ± 0.5	12.0	14.5	4.45 ± 0.20

Criterion

Regressions (slower)

Group	Bench	Param	Change	CI Lower	CI Upper
counts	single_property_indexed_entities		13.380%	12.021%	14.390%
sample_entity	sample_entity_single_property_indexed	100000	10.516%	9.620%	11.323%
counts	multi_property_unindexed_entities		5.928%	3.828%	8.044%
sample_entity	sample_entity_single_property_unindexed	10000	5.674%	1.921%	10.290%
counts	index_after_adding_entities		4.397%	3.993%	4.750%
sample_entity	sample_entity_single_property_indexed	1000	3.509%	2.134%	4.800%
algorithm_benches	algorithm_sampling_multiple_l_reservoir		3.317%	2.959%	3.652%
algorithm_benches	algorithm_sampling_multiple_known_length		2.868%	2.665%	3.078%
sample_entity	sample_entity_multi_property_indexed	10000	2.620%	1.371%	3.790%
indexing	with_query_results_single_indexed_property_entities		1.877%	1.330%	2.393%
sampling	sampling_single_known_length_entities		1.631%	1.305%	1.958%
indexing	query_people_count_single_indexed_property_entities		1.534%	1.297%	1.759%
counts	reindex_after_adding_more_entities		1.304%	1.065%	1.538%
indexing	query_people_count_multiple_individually_indexed_properties_enti		1.233%	1.061%	1.386%

Improvements (faster)

Group	Bench	Param	Change	CI Lower	CI Upper
sample_entity	sample_entity_single_property_unindexed	1000	-23.509%	-23.852%	-23.240%
indexing	query_people_single_indexed_property_entities		-19.706%	-20.024%	-19.266%
indexing	query_people_count_indexed_multi-property_entities		-19.535%	-19.803%	-19.115%
indexing	with_query_results_indexed_multi-property_entities		-13.616%	-14.382%	-12.999%
large_dataset	bench_query_population_derived_property_entities		-8.140%	-9.174%	-7.147%
large_dataset	bench_match_entity		-5.652%	-7.043%	-4.708%
sample_entity	sample_entity_multi_property_indexed	1000	-5.185%	-7.378%	-3.346%
sampling	sampling_single_l_reservoir_entities		-5.085%	-5.963%	-4.221%
large_dataset	bench_query_population_multi_indexed_entities		-4.641%	-5.061%	-4.298%
indexing	query_people_indexed_multi-property_entities		-4.190%	-4.425%	-3.945%
sampling	sampling_multiple_l_reservoir_entities		-2.572%	-2.722%	-2.442%
sampling	sampling_single_unindexed_concrete_plus_derived_entities		-1.995%	-2.236%	-1.748%

Unchanged / inconclusive (CI crosses 0%)

Group	Bench	Param	Change	CI Lower	CI Upper
large_dataset	bench_query_population_property_entities		-1.776%	-3.117%	-0.518%
sample_entity	sample_entity_single_property_indexed	10000	-1.394%	-2.181%	-0.691%
sample_entity	sample_entity_multi_property_indexed	100000	1.368%	0.341%	2.383%
examples	example-basic-infection		-1.204%	-1.694%	-0.745%
sampling	sampling_multiple_known_length_entities		-1.183%	-1.489%	-0.859%
sample_entity	sample_entity_whole_population	100000	0.739%	0.377%	1.198%
algorithm_benches	algorithm_sampling_single_l_reservoir		0.737%	0.547%	0.863%
examples	example-births-deaths		0.554%	0.276%	0.845%
indexing	with_query_results_multiple_individually_indexed_properties_enti		-0.525%	-0.973%	-0.066%
sampling	count_and_sampling_single_known_length_entities		-0.461%	-0.775%	-0.058%
large_dataset	bench_filter_unindexed_entity		-0.444%	-3.334%	2.767%
indexing	query_people_multiple_individually_indexed_properties_entities		-0.437%	-0.688%	-0.198%
sample_entity	sample_entity_single_property_unindexed	100000	0.395%	0.084%	0.706%
sample_entity	sample_entity_whole_population	10000	0.335%	-0.250%	0.919%
large_dataset	bench_filter_indexed_entity		0.330%	-7.247%	8.113%
large_dataset	bench_query_population_multi_unindexed_entities		-0.317%	-0.624%	-0.054%
sampling	count_and_sampling_single_unindexed_concrete_plus_derived_entiti		-0.166%	-0.238%	-0.090%
counts	multi_property_indexed_entities		0.155%	-0.073%	0.383%
sampling	sampling_multiple_unindexed_entities		-0.137%	-0.186%	-0.094%
algorithm_benches	algorithm_sampling_single_rand_reservoir		-0.137%	-0.306%	-0.016%
counts	single_property_unindexed_entities		-0.128%	-0.888%	0.586%
large_dataset	bench_query_population_indexed_property_entities		-0.109%	-0.317%	0.119%
sampling	sampling_single_unindexed_entities		-0.099%	-0.182%	-0.018%
counts	concrete_plus_derived_unindexed_entities		-0.076%	-0.350%	0.181%
algorithm_benches	algorithm_sampling_single_known_length		-0.057%	-0.507%	0.283%
sample_entity	sample_entity_whole_population	1000	-0.005%	-0.367%	0.253%

[k88hudson-cfa]

Looks good to me, it does make me wonder if this happens elsewhere in the codebase?