Use the public PyUnicodeWriter API

[vstinner]

Replace PyUnicode_New(), PyUnicode_CopyCharacters(), etc. with PyUnicodeWriter. It avoids creating incomplete Unicode strings: see Avoid creating incomplete/invalid objects.

Linked PRs

• gh-125196: Use PyUnicodeWriter in symtable.c #125199
• gh-125196: Add PyUnicodeWriter_Fill() function #125201
• gh-125196: Use PyUnicodeWriter for repr(list) #125202
• gh-125196: Add fast-path for int in PyUnicodeWriter_WriteStr() #125214
• gh-125196: Use PyUnicodeWriter for repr(structseq) #125219
• gh-125196: Use PyUnicodeWriter for repr(contextvars.Token) #125220
• gh-125196: PyUnicodeWriter_Discard(NULL) does nothing #125222
• gh-125196: Add a free list to PyUnicodeWriter #125227
• gh-125196: Use PyUnicodeWriter for repr(tuple) #125242
• gh-125196: Use PyUnicodeWriter for JSON encoder #125249
• gh-125196: Use PyUnicodeWriter in error handlers #125262
• gh-125196: Use PyUnicodeWriter for repr(dict) #125270
• gh-125196: Use PyUnicodeWriter in parser #125271
• gh-125196: Use PyUnicodeWriter in HAMT #125458
• gh-125196: Use PyUnicodeWriter in _io.StringIO #125528

[vstinner]

I also replaced PyUnicode_New(0, 0) with Py_GetConstant(Py_CONSTANT_EMPTY_STR) in commit b9a8ca0.