GeezerVison

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@crates/capi/src/unicodeobject.rs`:
- Around line 50-60: Update PyUnicode_FromObject to reject objects that are not
PyStr or PyStr subclasses instead of calling obj.to_owned().str(vm). Return the
existing C-API error representation for a TypeError, while preserving the
successful conversion path for accepted string objects.
- Around line 277-293: Update PyUnicode_CompareWithASCIIString to avoid calling
PyStr::expect_str(), which can panic for surrogate-containing strings. Compare
the PyStr value using its WTF-8 view, matching the behavior of
PyUnicode_Compare, while preserving the existing ordering-to-integer mapping and
UTF-8 validation for the C string.

---

Nitpick comments:
In `@crates/capi/src/unicodeobject.rs`:
- Around line 98-171: Extract the duplicated with_vm/downcast/encode_text logic
from PyUnicode_AsASCIIString, PyUnicode_AsLatin1String,
PyUnicode_AsRawUnicodeEscapeString, PyUnicode_AsUTF16String,
PyUnicode_AsUTF32String, and PyUnicode_AsUnicodeEscapeString into one private
encode_as helper accepting the encoding name, then have each exported ABI
function delegate to it with its existing literal while preserving all
#[unsafe(no_mangle)] symbols.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@crates/capi/src/unicodeobject.rs`:
- Around line 578-591: Update PyUnicode_Translate to honor the _errors parameter
by converting and passing it through to PyStr::translate, preserving the
requested codec error-handler behavior; if PyStr::translate cannot support it,
explicitly document the intentional API divergence instead.
- Around line 258-331: Update PyUnicode_DecodeUTF16 and PyUnicode_DecodeUTF32 to
honor the _byteorder pointer: when it is non-NULL, select the corresponding
forced little-endian or big-endian codec based on its input value, while
preserving BOM/native-order decoding when it is NULL. After decoding, write the
detected byte order back through the pointer as required by the API.