[refurb] Mark fix as unsafe when the right-hand side is a string (FURB171)

[InSyncWithFoo]

Summary

Resolves #15268.

Test Plan

cargo nextest run and cargo insta test.

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

ℹ️ ecosystem check detected linter changes. (+0 -0 violations, +0 -4 fixes in 2 projects; 53 projects unchanged)

zulip/zulip (+0 -0 violations, +0 -4 fixes)

ruff check --no-cache --exit-zero --ignore RUF9 --output-format concise --preview --select ALL

+ tools/lib/template_parser.py:637:29: FURB171 Membership test against single-item container
- tools/lib/template_parser.py:637:29: FURB171 [*] Membership test against single-item container
+ tools/lib/template_parser.py:645:29: FURB171 Membership test against single-item container
- tools/lib/template_parser.py:645:29: FURB171 [*] Membership test against single-item container

python-trio/trio (+0 -0 violations, +0 -0 fixes)

ruff check --no-cache --exit-zero --ignore RUF9 --output-format concise --preview

Changes by rule (1 rules affected)

code	total	+ violation	- violation	+ fix	- fix
FURB171	4	0	0	0	4

[AlexWaygood]

Nice, thanks! My only comment is that I wonder if we should even be emitting this lint if the r.h.s. is a string? It's quite possible that the user is deliberately exploiting the way in works with regard to string here, in which case it would be annoying to have to noqa the lint away.

That does seem unlikely, though, so maybe it's better to continue emitting the lint and just mark the fix as unsafe. Not sure -- what do you think?

[InSyncWithFoo]

I'm with the latter. In such cases, c == "" or c == "a" (or c in ("", "a") to avoid double evaluation) would be clearer and I would have suggested that as a secondary fix had it not been for the one-fix-per-diagnostic limit.

[AlexWaygood]

Thanks!