Raise a proper TypeError when a property assignment cannot be performed

[jhonabreul]

What does this implement/fix?

Fixes two related property-assignment bugs that surfaced together when a Lean Python algorithm assigned a pure-Python object to self.universe_manager (which snake-cases onto QCAlgorithm.UniverseManager { get; private set; }) and got the opaque SystemError: error return without exception set instead of an actionable error:

1. Converter.ToManagedValue could fail without setting a Python error even when called with setError: true. Three exits returned false silently: the final fall-through (plain Python object with no conversion to an arbitrary CLR class), the reflected-class-object case (converting a CLR class object to something other than Type), and non-MethodBinding managed types (method/module objects). PropertyObject.tp_descr_set trusts setError: true and returns -1, so CPython's release-build eval loop raises SystemError: error return without exception set at the assignment. These exits now set the same '{type}' value cannot be converted to {obType} TypeError the type_error label already uses, preserving any more specific error a probing sub-path (e.g. a throwing implicit operator) left pending.

2. Properties with private setters were silently writable from Python. PropertyObject.CacheAccessors resolved setters with GetSetMethod(nonPublic: true), so { get; private set; } properties bypassed the "property is read-only" guard and wrote engine-internal state through reflection (convertible values succeeded silently — the Lean repro replaced the algorithm's UniverseManager and hung the engine). The setter is now discarded when it does not satisfy ClassManager.ShouldBindMethod (the repo's accessibility policy: public, protected, protected internal), so private/internal setters read as read-only while protected setters keep working.

With both fixes, the Lean repro turns from SystemError: error return without exception set into TypeError: property is read-only, and non-convertible assignments to writable properties report TypeError: '<class>' value cannot be converted to <type>.

Each fix is a separate commit carrying its own regression tests; the tests were verified red before each fix and green after.

Does this close any currently open issues?

No open issue; root-caused from a Lean user report of error return without exception set during algorithm initialization.

Any other comments?

Behavioral note: code that (likely unknowingly) relied on assigning to private-setter C# properties from Python will now get TypeError: property is read-only. This aligns property writes with ClassManager.ShouldBindMethod, which already governs method/field exposure.

End-to-end verification against Lean master (the original repro): self.universe_manager = <pure python object> in a Python algorithm now fails with TypeError: property is read-only at the offending line instead of SystemError: error return without exception set, and a convertible value no longer silently replaces the algorithm's UniverseManager (which previously hung the engine on shutdown). Full embed suite on the combined branch: 974 passed, 0 failed, 8 pre-existing environment skips.