Feature/sep star detection

[jerrylewisbh]

The Problem
The existing star detection correlates each frame against a synthetic 15×15 blurred circle. This approach has some limitations on real allsky images: it assumes stars have a fixed shape that rarely matches a wide-angle lens, it lacks a sky background model (making it vulnerable to light pollution, clouds, and noise), and the threshold is a raw correlation value (0–1) with no physical meaning. The reported star count is often unreliable and difficult to tune.

Changes Made

• This PR adds SEP (Source Extractor Python) as an optional star detection method. SEP models and subtracts the varying sky background, then extracts sources based on a clear, physical sigma threshold (e.g., 5× the local noise).
• Implements the SEP detection algorithm as a drop-in option in the processing pipeline.
• Sizes the drawn marker circles proportionally based on the actual measured size of each star, so brighter/larger stars get larger markers.
• Updates the Settings UI with a new "Method" dropdown and dynamic threshold fields (enabling/disabling the appropriate threshold input based on the chosen method).
• Adds fallback handling for extraction errors and ensures the system handles mask conventions smoothly.
• Uses lazy loading for the new library, meaning the sep package is only imported and required if the user actively selects the SEP method.
• Backward compatibility: Preserves the existing template method as the default, ensuring no changes or disruptions for current users unless they explicitly switch.

Dependency Note: Requires pip install sep>=1.4.1 to use the new method.