Add resource-aware algorithm registration and execution

[knoepfel]

Phlex resources are framework-owned, user-defined objects that support algorithm execution without contributing data-flow edges or changing graph topology. Algorithms declare resource dependencies during registration; the framework supplies a resource access value for each invocation and, when needed, uses that value to limit concurrent access.

A resource type is ordinary user code. It need not derive from a Phlex base class or implement a common runtime interface.

Resource Access

Phlex defines three access categories:

1. Unlimited access: Any number of algorithm invocations may access the resource concurrently.
2. Single-token access: At most one algorithm invocation may hold the resource token.
3. Multi-token access: At most one algorithm invocation may run per available resource token.

Algorithms request resources with resource<T>{} after product selectors in input_family(...). The corresponding resource-access values are passed as trailing algorithm parameters.

oneTBB Integration

serial_node and serializer_node were Phlex prototypes used to explore resource-constrained execution. They are superseded by oneTBB's flow::resource_limiter and flow::resource_limited_node, which provide the scheduling mechanism adopted by Phlex.

Delivery Phases

Phase	Scope	Status
Phase 1 (#851)	Static registration; unlimited and single-token resources; resource-aware execution nodes	Implemented by PR #850, pending merge
Phase 2 (#852)	Resource registration from dynamically loaded plugins	Planned
Phase 3 (#853)	Multi-token resources and additional token-construction forms	Planned; awaits an official supported oneTBB release containing the required limiter constructors

Each phase issue owns its API details, implementation constraints, and acceptance criteria.

Not Currently Scheduled

• Mutable access to unlimited resources.
• Named resource instances or multiple instances of one resource type.
• Duplicate resource dependencies in one algorithm registration.
• Provider-node resources.