Previously, detectStackDriftRecursive returned raw CloudFormation API
outputs (flat maps of drift results, physical IDs, and detection IDs),
then drift.ts walked that same tree a second time in processData/
collectNestedStacks to fetch templates, compute counts, assign
categories, and produce a ProcessedDriftData object for the formatter.

Now detection builds an enriched StackDriftNode tree in a single pass:
each node carries its category, resource counts, and drift detection ID.
The formatter receives phase results directly and uses flattenTree for
uniform iteration. (also root node no longer handled separate, it is just the base case)

This reduces size of drift.ts (processData, collectNestedStacks,
and associated imports all removed), deletes the ProcessedDriftData/StackDriftData
intermediates, and moves count functions to detect-stack-drift.ts where
they belong.

 Additional changes:
  - Formatter receives (phase1, phase2, phase3, projectName) directly
  - Count functions and types moved from drift-formatter to stack cfn module
(this decouples the formatter, and now formatter no longer processes data) yay!

The API returns max 100 results per page. Previously a single call was
made, silently dropping resources beyond the first page. Now loops on
NextToken until all drift results are collected.

Also simplifies the return type from DescribeStackResourceDriftsCommandOutput
to StackResourceDrift[] since the wrapper was immediately unwrapped by
the only caller.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…yToAllowChange

Typed the propDiff parameter using the SDK's PropertyDifference type
instead of any, enabling compile-time validation of field accesses.

…o module scope

- Add PhysicalResourceId check before passing to ARN parser in
  analyzeChangeSet to satisfy TypeScript and guard against undefined
- Move stripAnsi from closure inside formatDashboardLine to module
  level to avoid re-creating it on every call

- countUnchecked now excludes nested stack logical IDs from the parent's
  unchecked count, since those are already handled by child nodes
- formatPropertyDiffs uses != null instead of truthiness so empty string
  values are displayed instead of silently omitted
- Convert require() calls to static imports in detect-local-drift
- Export ResourceChangeWithNested type from index barrel

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Keep only warn (consistent with gen2-migration logger and the underlying
amplify-prompts printer). Update the single call site in
detect-stack-drift.ts from print.warning() to print.warn().

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Convert CommonJS require calls to ES module imports for
CloudFormation provider class and S3BackendZipFileName constant,
enabling type checking and consistency with the rest of the codebase.

Was exported from both drift-detection/index.ts and
drift-detection/services/index.ts. Keep only the services barrel
since that's where the consumer imports from.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…drift

Static import of resource-status-data transitively loads
amplify-provider-awscloudformation which has top-level FeatureFlags
side effects that crash in test environments. Restore the lazy
require() pattern used elsewhere in the codebase (see amplify-toolkit.ts).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Wrap drift detection phases in an AmplifySpinner that updates as each
phase progresses. Printer methods pause/resume the spinner so log output
never collides with the animation.

Replace DriftSummary/ResourceCounts with a flat totalDrifted field on
CloudFormationDriftResults and a simple recursive countTotalDrifted.
Remove countInSync, countUnchecked, countFailed, countDrifted, and the
GetTemplateCommand call that only existed to support countUnchecked.