Finish GaussVonMisesDistribution implementation

[Copilot]

• Replace import numpy as _np with pyrecets.backend imports throughout gauss_von_mises_distribution.py
• Remove redundant asarray/array conversions
• Replace numpy imports with pyrecets.backend in test_gauss_von_mises_distribution.py
  • Removed import numpy as np and import numpy.testing as npt
  • Added allclose, squeeze to backend imports
  • Replaced np.random.rand with random.rand
  • Replaced np.zeros, np.linalg.solve, np.asarray, np.exp, np.cos, np.pi, np.squeeze with backend equivalents
  • Replaced npt.assert_allclose with self.assertTrue(allclose(...))
  • Replaced np.all(...) with all(...)
  • All 6 tests pass

[FlorianPfaff]

@copilot Do not use numpy, import from pyrecest.backend instead (as in other files).

[FlorianPfaff]

@copilot avoid use of asarray

[Copilot]

@copilot avoid use of asarray

Done in f1510ea. Removed all asarray usages — replaced with array() where conversion from scalars/lists was needed, and eliminated the unnecessary conversions on attributes that are already backend arrays (e.g. self.mu, self.P, self.A, self.Gamma, self.beta).

[FlorianPfaff]

@copilot don't use bumpy but rather pyrecest.backend even in the tests!