fix(py): make encode_with_unstable handle surrogates like encode (#541)

[LeSingh1]

Closes #541.

Encoding.encode and Encoding.encode_ordinary already catch UnicodeEncodeError raised by the Rust BPE when the input string contains surrogate pairs or lone surrogates, then retry after a UTF-16 surrogatepass + replace round-trip. Encoding.encode_with_unstable did not, so the same inputs that worked through encode raised UnicodeEncodeError here.

This PR mirrors the same try / except / repair pattern so the three encode paths agree on what inputs they accept.

Repro (against current main)

import tiktoken
enc = tiktoken.get_encoding("cl100k_base")

enc.encode("👍")               # works, returns [9468, 239, 235]
enc.encode_with_unstable("👍") # raises UnicodeEncodeError

Verification

Added test_encode_with_unstable_surrogate_pairs in tests/test_encoding.py. The test fails on main with the exact UnicodeEncodeError quoted in #541 and passes on this branch. I confirmed locally:

without fix: FAILED tests/test_encoding.py::test_encode_with_unstable_surrogate_pairs - UnicodeEncodeError
with fix:    PASSED tests/test_encoding.py::test_encode_with_unstable_surrogate_pairs

The new test cross-checks the stable prefix against enc.encode(...) to ensure both paths agree on the resulting tokens, not just that no exception is raised.

Scope

Single-file change in tiktoken/core.py (8-line try / except wrap mirroring encode's) plus one regression test.