Instrument transient GitHub user refresh failures

[jt-schmidt]

What

Adds targeted observability for transient GitHub network failures during Shipit::User#refresh_from_github!.

When a refresh fails with a known transient connection/timeout error (Faraday::ConnectionFailed, Faraday::TimeoutError, Net::OpenTimeout, Net::ReadTimeout), Shipit now:

• emits an ActiveSupport::Notifications event: transient_github_refresh_error.shipit
• logs a warning with non-email context: Shipit user id, GitHub id/login, exception class/message
• re-raises the original exception so existing retry/error behavior is unchanged

Why

PI investigation of shop/issues#16591 / Vault issue 15928 found the current Observe error group is from Shipit::RefreshGithubUserJob timing out while opening a TCP connection to api.github.com:443.

This keeps the first change low risk: add signal before changing retry/error reporting behavior.

Test plan

• Added coverage that Net::OpenTimeout during refresh_from_github! emits the notification, logs, and re-raises.
• ruby -c app/models/shipit/user.rb
• ruby -c test/models/users_test.rb

I could not run the full model test locally because the checkout is missing bundled native/test dependencies (Bundler::GemNotFound for sqlite3/ejson-rails/mysql2/etc.).

Refs: https://github.com/shop/issues/issues/16591