[ty] Fix subtraction overflow bug

[BurntSushi]

PR #21549 introduced a subtle overflow bug that seemed impossible, but
can empirically happen. This PR fixes it by saturating to zero.

I did try to write a regression test for this, but couldn't manage it.
Instead, I'll attach before-and-after screen recordings.

[BurntSushi]

Before:

2025-11-21T13.58.50-05.00.mp4

After:

2025-11-21T14.01.48-05.00.mp4

[astral-sh-bot]

Diagnostic diff on typing conformance tests

No changes detected when running ty on typing conformance tests ✅

[astral-sh-bot]

mypy_primer results

Changes were detected when running on open source projects

scikit-build-core (https://github.com/scikit-build/scikit-build-core)
- src/scikit_build_core/build/wheel.py:98:20: error[no-matching-overload] No overload of bound method `__init__` matches arguments
- Found 45 diagnostics
+ Found 44 diagnostics

No memory usage changes detected ✅

[astral-sh-bot]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

Formatter (stable)

✅ ecosystem check detected no format changes.

Formatter (preview)

✅ ecosystem check detected no format changes.

[Gankra]

Hmm I wonder if this is getting confused about utf8 vs utf16... 😨

[RasmusNygren]

Oh that was sloppy by me, not sure how I didn't see that in the playground. 🤔

This seems to (at the very least) happen when the cursor is in the middle of a token at the start of the file. This seems to reproduce it:

let builder = completion_test_builder(
            "\
f<CURSOR>oo = 1
",
);
builder.build();

[Gankra]

Oh I see, are we subtracting the size of the whole token from the cursor, rather than token-up-to-the-cursor?

[RasmusNygren]

We do get the range for the entire token

ruff/crates/ty_ide/src/completion.rs

Line 1266 in 438ef33

Some(source[last.range()].to_string())

The subtraction is probably a bit off in general then and there might be some other scenarios where we don't suppress suggestions (even though we should) if the cursor is not at the end of the current token. Had a quick check but couldn't figure out a test case that trigger such a scenario even though I feel like they must exist. I'll have a think.