Process smartstack frames in AWS banzai (e45 --> e91)

[timbeccue]

This change configures banzai to process the smartstack fits files (e45) that are sent to the archive, producing a final e91 as output.

To recap the current smartstack process:

• site banzai: receives n00 (raw stackframes)
• site banzai: reduce the n00 frames to produce n09 stackframes (using the same complete reduction stages as EXPOSE frames)
• site banzai: stack the n09 frames into one e45, and send to the archive
• AWS banzai: receives the e45 and do another reduction pass to create an e91 frame

The stages run on the e45 frame are:

• SourceDetector
• WCSSolver
• PointingTest
• PhotometricCalibrator

This PR also includes a version bump from 1.38.1 to 1.38.2.

[cmccully]

Looks ok to me.

[Copilot]

🟢 Approval recommended

The routing, stage selection, naming behavior, documentation, and version metadata are consistent and adequately tested.

Pull request overview

Enables central BANZAI to accept Smartstack e45 frames and produce e91 archive products.

Changes:

• Routes configured reduction levels to a specific starting pipeline stage.
• Updates output naming and adds routing/stage-selection tests.
• Documents the workflow and bumps the release to 1.38.2.

File summaries

File	Description
banzai/settings.py	Maps e45 inputs to SourceDetector.
banzai/utils/stage_utils.py	Supports starting midway through the pipeline.
banzai/utils/realtime_utils.py	Accepts configured nonzero reduction levels.
banzai/lco.py	Rewrites terminal reduction-level digits safely.
banzai/tests/test_stage_utils.py	Tests stage selection and execution.
banzai/tests/test_need_to_process_image.py	Tests archive routing by reduction level.
banzai/tests/test_frames.py	Tests e45 to e91 filenames.
docs/smartstacking_architecture.md	Documents central tail reduction.
CHANGES.md	Records the 1.38.2 feature.
pyproject.toml	Bumps the package version.
uv.lock	Synchronizes the locked project version.

Review details

• Files reviewed: 10/11 changed files
• Comments generated: 0
• Review effort level: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.