refactor!: migrate to modern numpy API, patch root solving, and bump dependencies

- Drop support for Python 3.8.
- Bump minimum dependencies to `numpy>=2.0` and `numba>=0.60`.
- Replace legacy NumPy polynomial functions (polyder, polyadd, etc.) with
  the modern `numpy.polynomial.Polynomial` class for stable calculus and
  arithmetic operations.
- Remove `_strip_leading_zeros` as the modern API handles degree alignment
  and zero-trimming natively.
- Retain highest-to-lowest `self.coefficients` array to maintain compatibility
  with Numba and CUDA GPU kernels evaluating via Horner's method.
- Fix `TypeError` in `quadratic_solve` by evaluating `b.real >= 0` to safely
  support complex coefficients.
- Update `quadratic_solve` return type hint to `np.ndarray`.

.gitea/workflows/test.yml

@@ -6,7 +6,7 @@ jobs:
     runs-on: nvidia-cuda
     strategy:
       matrix:
-        python-version: ["3.8", "3.10", "3.12", "3.14"]
+        python-version: ["3.10", "3.12", "3.14"]
 
     steps:
     - name: Checkout repository