feat: Group pixi-update lockfiles into one pull request

[Yannik Tausch (ytausch)]

Motivation

Closes #30.

quant-ranger currently opens one pull request per updated pixi.toml. In
monorepos with many manifests this produces a lot of pull requests, and merging
them triggers repeated CI runs. As suggested in the issue, each manifest can now
declare a group in its [tool.update] section, and all manifests with the same
group are published in a single pull request.

Changes

• Add an optional group key to [tool.update] in pixi.toml. It is
  branch-safe (^[A-Za-z0-9][A-Za-z0-9._/-]*$) because it is used as a branch
  segment.
• Lockfiles of the same repository that share a group are regenerated in one
  checkout and published together as one pull request. The branch is
  <autoupdate-branch-prefix>/<group>, the title is
  <autoupdate-commit-message> (<group>), and the body contains one section per
  changed lockfile.
• Grouped manifests must agree on autoupdate-branch-prefix,
  autoupdate-commit-message, and autoupdate-pull-request-labels; a
  disagreement fails the group instead of silently picking one manifest's values.
• Lockfiles that are already up to date are not staged and are not listed in the
  body. A failing lockfile only fails its own item; the remaining lockfiles are
  still published.
• The updater base gains two small seams: _update_units groups items that share
  a run, and _run_unit executes one group, so one updater run can publish
  several items through a single checkout while still emitting one result per
  item. Outcome recording is shared between the normal and unexpected-failure
  paths. Default behavior (one checkout and one pull request per item) is
  unchanged.
• Document the option in the pixi-update reference and the managed-branches table.

Notes:

• group is per repository, so it does not introduce a new config file and stays
  extensible. Groups are keyed by repository and group name.
• The dependency report of each changed lockfile is still produced with
  pixi-diff-to-markdown and concatenated under per-manifest headings. Filtering
  for major upgrades/downgrades in pixi-diff-to-markdown remains a separate
  follow-up, since that is upstream work.
• Posting one review per pixi.lock (instead of the body table) is not part of
  this change; the current body-table behavior is kept for both grouped and
  single-lockfile updates.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (3f5379d) to head (523dd36).

Additional details and impacted files

@@            Coverage Diff             @@
##              main       #35    +/-   ##
==========================================
  Coverage   100.00%   100.00%            
==========================================
  Files           49        49            
  Lines         3077      3177   +100     
==========================================
+ Hits          3077      3177   +100     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.