Instructor onboarding

[shaidar]

Description (What does it do?)

Setup dagster pipeline for instructor onboarding using our existing access-forge process to add users with associated roles.

[Copilot]

Pull Request Overview

This PR implements a Dagster pipeline for instructor onboarding by automating the process of extracting user email addresses from course role data and submitting them to the access-forge GitHub repository for access management.

Key Changes:

• Creates a GitHub API client resource for Dagster pipelines with Vault-based authentication
• Implements two new assets: one to generate a CSV of unique user emails from course roles, and another to create GitHub pull requests with this data
• Integrates the new assets into the existing edxorg daily job schedule

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
packages/ol-orchestrate-lib/src/ol_orchestrate/resources/github.py	Adds a new GitHub API client factory resource that retrieves credentials from Vault
dg_projects/edxorg/edxorg/definitions.py	Registers the GitHub resource and new instructor onboarding assets with the Dagster pipeline
dg_projects/edxorg/edxorg/assets/instructor_onboarding.py	Implements assets for extracting user emails and creating PRs to the access-forge repository

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

The user count calculation is incorrect. This counts newline characters in the CSV string, which includes the header row. The actual user count should be instructor_onboarding_user_list.count(chr(10)) - 1 to exclude the header, or better yet, pass the count from the previous asset's metadata.

Suggested change

	f"- Users: {instructor_onboarding_user_list.count(chr(10))} entries"
	f"- Users: {instructor_onboarding_user_list.count(chr(10)) - 1} entries"

[Copilot]

The cached client instance in _client is not thread-safe. If multiple Dagster ops/assets access this resource concurrently, they could encounter race conditions during client initialization or reuse a client with stale state. Consider using a lock or removing the cache to create a new client on each call.