Audit and fix legacy bondi_hoyle implementation bugs

Context

While extracting reusable XRB calculations for #881, several suspicious behaviors were identified in posydon.utils.common_functions.bondi_hoyle. They should be investigated and fixed in a dedicated PR rather than folded into the new-module PR.

Items to investigate

1. A variable named he_core_mass is populated from he_core_radius and then implicitly converted to a Boolean.
2. The slow-wind assignment for extended H-rich giants (beta = 0.125) is followed by mass-based assignments that appear to overwrite it for every valid positive stellar mass.
3. In the Kudritzki+2000 branch, f_m is assigned as a scalar inside a loop, so an array calculation can apply the last system's factor to all systems.
4. An unsupported wind-scheme string can reach the calculation with f_m undefined instead of raising a clear ValueError.
5. The random single-phase treatment and the appropriate orbit-averaged alternative require the physics review already discussed in #829.
6. The wind-disc criterion mixes velocities calculated in SI with const.clight in cgs through an implicit 0.01 conversion, which should be made explicit and tested.

Acceptance criteria

• Confirm the intended prescriptions from the cited literature before changing behavior.
• Separate definite implementation bugs from choices of physical approximation.
• Add scalar and vector regression tests for each confirmed bug.
• Test H-rich giants, He-rich donors, mixed-temperature arrays, invalid schemes, circular/eccentric systems, and wind-disc outcomes.
• Keep the repair separate from #881.