Don't add unnecessary parentheses during inline variable extraction

[ak4-sh]

Summary

Fix inline variable refactoring to only add parentheses when they are needed for the expression being inlined and the context where it is inserted.

• Parenthesize expressions that require grouping, such as binary operations, boolean operations, comparisons, tuples, lambdas, and similar expression forms.
• Avoid unnecessary parentheses for simple values like number literals in normal expression contexts.
• Add parentheses for number literals when inlining into attribute access, e.g. value = 42; value.bit_length() becomes (42).bit_length().

Fixes #3882

Test Plan

Added LSP code action tests for:

• Binary expression inlined into attribute access.
• Number literal inlined without unnecessary parentheses.
• Number literal inlined into attribute access.
• Boolean operation requiring parentheses.
• Tuple expression requiring parentheses.

[NathanTempest]

Thanks for the fix, @ak4-sh !

One tiny optional follow-up (not blocking): NumberLiteral also matches floats/complex, but only integer literals have the 42. ambiguity
4.2.foo and 4j.foo are already valid Python, so they'll get a harmless-but-redundant (...). If you want to be minimal you could narrow that check to integer literals, but the current behavior is perfectly safe. Thanks again!

[meta-codesync]

@NathanTempest has imported this pull request. If you are a Meta employee, you can view this in D110088158.

[kinto0]

Review automatically exported from Phabricator review in Meta.

[ndmitchell]

Review automatically exported from Phabricator review in Meta.

[ak4-sh]

Thanks for the comment @NathanTempest! I've updated the match condition, do let me know if it looks okay

[NathanTempest]

Good work! The Number::Int narrowing is exactly right, floats/complex no longer get redundant parens while decimal ints (42.bit_length()) are still handled. One small convention nit: we keep imports at the top rather than inline-qualified paths, so use ruff_python_ast::Number; + Number::Int(_) (matches how alt/expr.rs does it). Optionally, a quick negative test that 4.2.foo stays unparenthesized would lock in the new behavior. Thanks for iterating on this!

[ak4-sh]

@NathanTempest Made the updates, sorry about this!