Add convex hull function

[linwang9926]

• Add get_convhull function.
• Test the function using canola and rice models, instance(s) with NaN nodes.

[codecov]

Codecov Report

Merging #13 (ecca898) into main (6faba20) will increase coverage by 1.32%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main      #13      +/-   ##
==========================================
+ Coverage   93.20%   94.53%   +1.32%     
==========================================
  Files           6        7       +1     
  Lines         103      128      +25     
==========================================
+ Hits           96      121      +25     
  Misses          7        7              

Impacted Files	Coverage Δ
sleap_roots/convhull.py	100.00% <100.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[talmo]

Small change to factorize the function

[talmo]

This is maybe a semantic preference, but I'd call this get_convhull_features.

The convex hull is technically defined by a set of points, so that's what I'd expect to get back from calling a function named get_convhull.

A better organization here might be to split this into two functions, where one just gets the convhull points:

def get_convhull(pts: np.ndarray) -> Optional[ConvexHull]:
    pts = pts.reshape(-1, 2)
    pts = pts[~(np.isnan(pts).any(axis=-1))]

    if len(pts) <= 2:
        return None

    # Get Convex Hull
    hull = ConvexHull(pts)

    return hull

And the second one which takes the ConvexHull instance and gives you back features. I appreciate it's more cumbersome to have to call two functions, so you could add a convenience checker to compute the convhull if needed:

def get_convhull_features(pts: Union[np.ndarray, ConvexHull]):
    hull = pts if type(pts) == ConvexHull else get_convhull(pts)

    if hull is None:
        return np.full((7,), np.nan)
    
    # perimeter
    perimeter = hull.area
    ...

[talmo]

LGTM