refactor(task-prioritization): extract service object, fix bugs, add tests

[Niethin69]

Description

This PR stacks on top of #94 to address review-stage issues found in the original implementation. It refactors the scoring logic into a dedicated service object, fixes correctness bugs in the task status filter and nil handling, resolves an N+1 query, and adds unit and API test coverage.

Please review #94 first. This PR is best merged after #94 lands; until then, GitHub will show @rashi-agrawal29's commit in the diff alongside mine.

Refs #94

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Summary of changes

Refactor

• Extracted scoring logic from TaskPrioritizationApi into a new TaskPrioritizationService under app/services/. The API file is now ~15 lines and matches the auth → service → present convention used elsewhere in the codebase.
• Promoted magic numbers (deadline buckets, effort buckets, scoring weights, task pressure thresholds, target grade scores) to named constants at the top of the service.
• Exposed effort_score_for(task) as the integration seam for the upcoming AI Task Prioritisation Service.

Bug fixes

• Status filter: the original where.not(task_status_id: 2) was fragile (relied on seed order) and incomplete (did not cover discuss or demonstrate, both of which mean "no further student action required"). Now resolved by name via TaskStatus.where(name: %w[complete discuss demonstrate]).
• N+1 queries: replaced .joins with .includes for task_definition and project => unit, both of which are accessed inside the result-building loop.
• Nil handling: Project.average(:target_grade) returns nil for a student with no enrolments; the previous .to_f.round coerced this to 0 and incorrectly returned the Pass score. Now falls through to an explicit default. task_definition.weighting is now coalesced explicitly. Removed misleading current_user&.id since authenticated? guarantees non-nil.

How Has This Been Tested?

Added two new test files:

• test/api/task_prioritization_service_test.rb — covers deadline scoring buckets (including overdue and nil due-date), effort buckets, sort order, filtering by user / unit activity / enrolment / task status, response schema, and workload scoring.
• test/api/task_prioritization_api_test.rb — covers authentication (401 unauthenticated), empty response, sort order, and response schema.