feat: add --order option to run tests in random order

[kirthi-b]

PR Checklist

• Addresses an existing open issue: fixes 🚀 Feature: --order option for random test order #902
• That issue was marked as status: accepting prs
• Steps in CONTRIBUTING.md were taken

Overview

Implements the --order option from #902, following the design approved in
#902 (comment) and the
option shape from the issue body: --order default, --order random and
--order random:<seed>.

How it works:

• Sibling suites and sibling tests are shuffled (Fisher-Yates) within their
  parent suite, per 🚀 Feature: --order option for random test order #902 (comment).
  The hierarchy is preserved, so before/after/beforeEach/afterEach run
  exactly as often as in a normal run.
• Every randomized run reports its seed: Mocha: test order randomized with seed 42; use --order random:42 to reproduce. The seed goes to STDERR, so
  machine-readable reporters on STDOUT (json, xunit) are unaffected. This was
  the main open question on the old reference PR feat: randomization support [WIP REFERENCE - DO NOT MERGE] #2301.
• The shuffle happens in Runner#run's prepare step, so it also works with
  --delay and inside parallel workers. In parallel mode the main process
  bakes the generated seed into the options before they are serialized, so all
  workers shuffle from the same seed and the seed is printed once.
• --sort and --order random error out as mutually exclusive, same as the
  existing --parallel and --sort pairing.
• Also available programmatically via mocha.order('random:42') and the
  order constructor option.

Includes unit tests, integration tests (fixed seed order, reproduce-from-
reported-seed round trip, error cases) and docs for the new flag.

Notes for reviewers:

1. The 2024 comment suggests targeting the randomization branch, but that
   branch no longer exists on the remote, so this targets main.
2. RNG: the approval said seedrandom "or an equivalent". This uses a small
   inline seeded PRNG (mulberry32) to avoid adding a runtime dependency;
   identical seeds give identical orders on any platform. Happy to switch to
   seedrandom if you prefer the dependency.
3. Seed reporting currently writes to STDERR (console.log in browsers). If you
   would rather surface it through reporters or a runner event, I can rework
   that.
4. Only random and default are implemented; the random-suites variant
   floated mid-thread was left out to keep scope to the approved design.

Closes #902

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: kirthi-b / name: kirthi_b (de942cc)

[mark-wiemer]

Looks OK, pretty big and low-priority, but I'll add it to my backlog :)

[mark-wiemer]

FYI @JoshuaKGoldberg the backing issue has 80+ likes!