Add work provider signals for activity eligibility in workflows.

Author: bpotchik

docs/dev/workflows.md

@@ -84,6 +84,8 @@ Activities can be configured with `continuation` eligibility, allowing them to:
   - Execute repeatedly until becoming ineligible
   - Build iterative processing patterns
 
+Continuations can derive their eligibility from work providers, which are Activities that report available work through their eligibility handlers. When a continuation has work providers, it remains eligible as long as any provider reports work, even if the continuation’s own handler does not. This allows Activities to signal that additional work is needed without requiring the continuation to directly poll or inspect external state.
+
 ### Workflow Machine
 
 The **Workflow Machine** is the engine that manages and executes a Workflow. It controls the execution flow according to the Workflow's blueprint, handling states, commands, and activity eligibility. The Workflow Machine ensures that analyses are performed in the correct order and manages state transitions during the analysis process. We have built an API to interact with the Workflow Machine, allowing you to issue commands, query and modify state, and control activity eligibility.
@@ -157,7 +159,7 @@ Additionally, some Activities currently act as containers for multiple analysis
 
 ### Workflow Cloning and Registration
 
-Binary Ninja utilizes a copy-on-write mechanism to enable safe workflow customization. This approach allows you to create and modify workflows without impacting the integrity of registered workflows that may already be in use. When you clone a workflow, you generate a new, independent version that can be safely customized. Once the cloned workflow has been tailored to your needs, it can be registered under a new unique name, or—if the original workflow is mutable—under the same name to seamlessly replace it.
+Binary Ninja uses a copy-on-write model to make workflow customization safe. When you clone a workflow, you create an independent copy that can be modified without affecting the original or any analyses already using it. After customizing the clone, you can register it under a new name or, if the original workflow is mutable, register it under the same name to replace it.
 
 The process involves three steps:
 
@@ -331,7 +333,7 @@ The Activity's callback function operates on an `AnalysisContext`, which provide
 modifying, or transforming the context.
 
 - **Workflow-Defined Composition**:
-The structural relationship of an Activity—whether it has children or serves as a child of another Activity—is defined by the Workflow. The Workflow determines the execution order and dependency graph, making Activities composable and adaptable to various analysis pipelines.
+Activities do not define their own hierarchy. Instead, the Workflow determines parent/child relationships, execution order, and dependencies. This allows Activities to be reused and composed into different analysis pipelines.
 
 - **Declarative Metadata via JSON Configuration**:
 	Each Activity is accompanied by metadata that provides essential information for its integration and execution within a workflow. The metadata structure ensures that Activities are not only functionally robust but also easy to manage and integrate into complex workflows. This metadata is defined through a JSON configuration and serves several purposes:
@@ -343,7 +345,7 @@ The structural relationship of an Activity—whether it has children or serves a
 	The role field specifies the Activity's operational role within the workflow, such as `action`, `task`, or `subflow`. This helps determine how the Activity interacts with others in the workflow.
 
 	- **Eligibility and Execution Semantics**:
-	The eligibility field defines conditions under which the Activity is allowed to run. It supports a range of predicates, including automatic settings, single-run constraints, and custom logical operators, allowing for fine-grained control over execution.
+	The eligibility field defines conditions under which the Activity is allowed to run. It supports a range of predicates, including automatic settings, single-run constraints, work provider signals, and custom logical operators, allowing for fine-grained control over execution.
 
 ### Activity Configuration
 
@@ -382,7 +384,8 @@ Eligibility determines whether an Activity should execute, based on certain cond
 	- `default`: The default boolean value for the auto-generated setting. If not provided, the default is `true`.
 - `runOnce`: A boolean indicating whether the activity executes only once across all file/analysis sessions. Once the activity runs, its state is saved persistently, and it will not run again unless explicitly reset. This is useful for activities that only need to be performed once, such as initial setup tasks.
 - `runOncePerSession`: A boolean indicating whether the activity executes only once within the current session. Its state is not saved persistently, meaning it resets when a new session begins. This is useful for tasks that should run once per analysis session, such as initialization steps specific to a particular execution context.
-- `continuation`: A boolean indicating if a `subflow` is eligible for re-execution based on its eligibility logic. If not provided, the default is `false`.
+- `continuation`: A boolean indicating if an activity is eligible for re-execution based on its eligibility logic and/or work provider signals. When enabled, the activity remains on the execution stack after processing and is re-evaluated each cycle. If not provided, the default is `false`.
+- `workProvider`: A boolean indicating that this activity signals whether it has work available via its eligibility handler. Work providers are evaluated by ancestor continuation activities to determine if additional processing cycles are needed. Not compatible with the `selector` role. If not provided, the default is `false`.
 - `predicates`: An array of objects defining the condition that must be met for the activity to be eligible to run.
 	- `type` (required): A string indicating the type of predicate. Valid values:
 		- `"setting"`: Evaluates the value of specific setting.