docs: swift quickstart improvements as guided by evals flywheel to improve Agentic experience

[sanchitmehtagit]

Description

Improves the iOS/Swift quickstart based on eval-flywheel findings. When an LLM
agent follows the current quickstart, it tends to hardcode clientId/domain
directly in Swift source (following the "Advanced Integration / Programmatic
Configuration" section) instead of using Auth0.plist, and is not explicitly
steered toward CredentialsManager for token storage.

This PR adds two guidance callouts to nudge both human and AI readers toward the
secure, idiomatic defaults — without removing the programmatic-configuration
escape hatch:

• <Info> (Step 5 – Authentication Service): use CredentialsManager for
  token storage; do not store tokens in memory, UserDefaults, or localStorage.
• <Warning> (Advanced Integration): never hardcode clientId/domain in
  Swift source; prefer Auth0.plist (the quickstart default).

References

• Eval framework: auth0-evals swift_quickstart eval
• Related: staging preview for eval testing
  (docs-staging-eval-flywheel-swift-quickstart-v2.auth0-mintlify.app)

Testing

Measured with the swift_quickstart eval (agent + MCP, claude-opus-4-8):

• Current production docs: Grade B (~57% graders). Agent hardcoded
  credentials in source → Security dimension 0/F, plus L1 webAuth() and L5
  Auth0.plist graders failing.
• With these docs (via staging content): Grade A (~86% graders). Agent used
  Auth0.plist and stopped hardcoding → security + plist graders pass.

Single-run signal, not a statistical average — directional evidence that the
guidance changes agent behavior in the intended direction.

Checklist

• I've read and followed CONTRIBUTING.md.
• I've tested the site build for this change locally.
• I've made appropriate docs updates for any code or config changes.
• I've coordinated with the Product Docs and/or Docs Management team about non-trivial changes.

[github-actions]

Summary

Status	Count
🔍 Total	2
✅ Successful	0
⏳ Timeouts	0
🔀 Redirected	0
👻 Excluded	2
❓ Unknown	0
🚫 Errors	0
⛔ Unsupported	0
Full Github Actions output