Fix slice repr leaking the internal storage representation.

[jseop-lim]

Description

Fixes #1071

slice.__repr__ formatted its three members with Java's String.valueOf instead of Python repr(), so any member whose debug toString() differs from its repr leaked the internal representation. int and None members were unaffected, which is why the common case looked correct.

AS-IS

slice(()), slice([]), slice(""), slice(True, 2):

slice(None, tuple(EmptySequenceStorage[]), None)
slice(None, list(EmptySequenceStorage[]), None)
slice(None, , None)
slice(true, 2, None)

TO-BE

slice(None, (), None)
slice(None, [], None)
slice(None, '', None)
slice(True, 2, None)

(matching CPython)

Changes

• Rewrite SliceBuiltins.ReprNode to apply PyObjectReprAsTruffleStringNode to getStart() / getStop() / getStep() and assemble the result with SimpleTruffleStringFormatNode, matching CPython's slice(%R, %R, %R). RangeBuiltins.ReprNode already uses this shape for range, whose CPython repr formats its three members with %R in the same way. The specialization no longer needs @TruffleBoundary. Unlike range, whose members are always integers, a slice member can be an arbitrary object with a user-defined __repr__, so the repr node is called with the frame rather than null.
• Add test_repr to test_slice.py, covering the member types whose Java toString() diverges from their repr, and an object with a user-defined __repr__ to pin that the member is formatted by Python repr().

Testing

mx python-jvm on linux-aarch64, then mx graalpytest test_slice.py test_list.py test_tuple.py test_bytes.py test_memoryview.py test_range.py — 224 tests, all pass. repr() of 15 slice forms (one-, two- and three-argument, with int / None / bool / float / str / bytes / tuple / list / dict members and an object with a user-defined __repr__) was compared against CPython 3.13.5 and matches on every one.