Adaptive Queue V2

[dxenes1]

Previously handled through ad-hoc notebook and custom scripting, now an official feature on NeuVue

Summary
This branch adds optional server-side adaptive priority scoring for tasks. Tasks can now carry an adaptive_priority object using the weighted_sum_v1 strategy, allowing the API to compute scalar priority from numeric or categorical priors, weights, bias, min/max clamps, and provenance metadata.

Changes

• Added adaptive_priority schema support on tasks, including priors, weights, categories, bias, score, contributions, provenance, and updated.
• Added src/utils/adaptive-priority.ts for normalization, validation, weighted score calculation, shallow patch merging, and task payload application.
• Added PATCH /tasks/{taskId}/adaptive_priority to merge adaptive priority inputs and recompute scalar priority when enabled.
• Updated task creation so adaptive tasks compute priority before Mongoose validation, while existing scalar-priority tasks remain backward compatible.
• Added a task index for adaptive queue reads: { active, status, namespace, priority }.
• Updated OpenAPI docs with the new AdaptivePriority schema, endpoint, examples, and queue query guidance.
• Refactored controller update paths toward promise-based Mongoose APIs and shared update error handling.
• Updated Restify/Mongoose/runtime dependencies and moved the project to Node >=20.19.0.
• Updated Docker local dev/runtime setup to Node 20, MongoDB 8, health checks, .dockerignore, and host port 9005.

Testing

yarn run build: passed
yarn run unit-test: passed, 18 tests
yarn run tslint: currently fails on existing project-wide lint issues unrelated to the adaptive priority change.