Refactor: dedicated downloader subsystem

[RBhupi]

Area

Acquisition — data download or queuing

Motivation

Developer experience — improves development or testing

Problem

The current NEXRAD pipeline relies on the external nexradaws package for Level II data acquisition. While it works well, it introduces following limitations:

• not available through conda-forge
• largely unmaintained
• feature requests and bug fixes cannot be addressed upstream
• downloader implementation is tightly coupled with the pipeline
• difficult to extend to realtime monitoring or alternative data sources

Proposal

Introduce a new top-level package

adapt/
    downloaders/
        base.py
        registry.py
        nexrad/
            downloader.py
            aws.py
            watcher.py
            models.py
            filters.py

Downloaders become reusable infrastructure components rather than scientific modules.

Proposed solution

The first implementation will be a lightweight native NEXRAD AWS downloader built directly on boto3.

Goals

• remove dependency on nexradaws
• direct boto3 implementation
• support archive and realtime workflows

Future CLI

adapt run-nexrad \
    --only download \
    --start 2026-05-01T00:00 \
    --end 2026-05-01T06:00 \
    --radar KLOT

which downloads data without executing the scientific pipeline.

Alternatives considered

No response

Additional context

No response

Checklist

• I searched existing issues and this has not been requested before
• This is one focused request, not multiple features bundled together