enh: suppress false-positive unknown-name errors after short-circuit boolean ops

[MarcoGorelli]

Summary

Currently, in a check like

if SOMETHING_DEFINITELY_FALSE and A:
    pass

, unknown-name is raised for A, even though it's not reached. This came up in Polars, where there's a variable defined based on the Polars version, but

if sys.version_info >= (3, 15) and A:

raises

The idea is that, if a boolean expression short-circuits based on an always-true statement (if the boolean expression is or) or an always-false one (if the boolean expression is and), then the code after it is unreachable and so unknown-name shouldn't be reported

The check for short-circuiting needs to be done:

• for the first value in the boolean op (values.next)
• the subsequent values (when iterating over values)

AI disclosure: I used some assistance from Claude

Fixes #3554

Test Plan

This PR includes tests. The mypy-primer also shows no effects

[github-actions]

According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅